1) China can (and does) use the models to influence the west. They train in false information about Taiwan and Hong Kong. Or pretend like history is in favor of China.
2) Ignoring the models containing false information, they are incredible. But you should be scared of running inference via the model creators directly. If you think your data is safe compared to running it via model providers in the US ( either frontier or model hosts like fireworks.ai ) then please let me know your bank details so I can poke around.
Agree, China believes in Climate change and are at least taking steps to address it. Personally this is making me trust them more than the USA because, facts.
I mean lesser of two evils thinking, if one is intentionally leading us towards climate disaster, while the other isn't then yeah. What else can be said? Should I trust the authoritarian country who believes in engineering and science, or the one that doesn't?
1) I'm not using AI to bicker over fringe political shibboleths.
2) I don't think it is any more or less safe to put my code on a Chinese server versus an American one. A Chinese provider also isn't liable to spy on me for the feds, as OpenAI and Anthropic certainly do.
Regarding point 2, I don't trust my data being safe running inference on model creators api, but neither do I trust US providers. Both use it for their own benefit, the only difference is the country of origin. The US has a lot more legal safeguards for this but I don't trust they don't do it regardless.
legal safeguards only make sense only when they’re enforced. With how “move fast and break things” Silicon Valley is law is always playing catch up (at your expense)
Why would it be a problem that they're Chinese? It's a problem because their country is ruled by an authoritarian regime. The West, by contrast, doesn't have a singular arbiter of truth, it has a plurality of perspectives (as much as certain interest groups would really rather that was not the case). It's not perfect, but those of us who have memories of living under authoritarian regimes can tell you there's no comparison.
On (1), models are an aggregation of large volumes of data sources, whatever the culture producing them, you'll get the average bias of that culture.
We see that on what minorities are associated with inside the model, or how things that aren't online will have a completely different weight. Or how 2/4/5/8ch or X will be disproportionately present in specific models despite being the places where facts go to die.
I see your point, but this isn't exactly true, it only takes a single person to bias a model by deleting specific training data or over training on certain facts.
But hasn’t it been the same with US since the WWII? US has influenced the world through various ways sometimes even weaponising human rights to spread American superiority and its narrative?
While true, you neglect the fact that US corporations have some autonomy from the government. I doubt that any of the current American AI corporations get regular dossiers of desirable responses to sensitive political topics from Washington and implement them. That would be a reason for a major lawsuit and a major scandal.
In China, the image of the country and the preferred narrative is under much tighter control of the government and local AI shops won't have any autonomy in this regard at all. Either obey or get shut down.
BTW What you mean by "weaponizing human rights", exactly? I am curious. If anything, I would say that the US foreign policy didn't promote human rights sufficiently, especially in Latin America, where the "bastard, but our bastard" attitude was typical.
OTOH in Europe, US human rights policy was probably relevant in saving some dissenters in the former Eastern Bloc from torture or execution.
> They train in false information about Taiwan and Hong Kong. Or pretend like history is in favor of China.
American labs can open their models or their old models at any point if they actually care about this, but until the day OG GPT 4 isn't averrable to download or Claude 3 then they're only pretending to care about this because they can profit from restrictions.
It's incredible how accurate this is towards US models:
> The 2 things people need to remember:
> 1) USA can (and does) use the models to influence the rest of the planet, and put political pressure. They train on stolen data, hide information, gate keep, who knows what they're hiding. In favor of the USA, nevertheless.
> 2) Ignoring the models containing false information, they are incredible. But you should be scared of running inference via the model creators directly. If you think your data is safe compared to running it via model providers in China ( either frontier or model hosts like z.ai ) then please let me know your bank details so I can poke around.
There's an irony in it. I have many reasons to believe China has its own geopolitical narratives and goals just like the US does. If someone dislikes when the US does it, why ignore when China does the same thing and praise them for their efforts? These models exist and sure, they accomplished something interesting. Use them with caution but I don't appreciate the political narratives about good and bad guys and AI models and politics.
> 1) China can (and does) use the models to influence the west. They train in false information about Taiwan and Hong Kong. Or pretend like history is in favor of China.
One of the interesting things is that through a fairly rudimentary process which is being done by 3rd party amateurs who've downloaded the open models, models like Qwen 3.6 35B-A3B (or 27B) can be fully 'uncensored' when turned into GGUF files.
I have an uncensored Q8 version of Qwen 3.6 35B-A3B here that will very happily output information about Tiananmen Square, Uyghurs, human rights in China, or indeed can even be instructed to write an intentionally absurd vitriolic screed against the CCP. The same uncensored 27B (dense) will do the same, just at a slower token/s rate.
Similarly there's 'uncensored' variants of Gemma4 31B and other western trained models, which once put through the same process, will also discuss or write just about anything you want, bypassing whatever internal guard rails were attempted in the training data set.
edit: more concerning, and a very legit concern, is that a model is only as good as the sum total of its training dataset, so if something is trained on a steady diet of news sources like Peoples Daily, Xinhuanet and similar in the English language, then it'll have a greater percentage of CCP-approved media publications in its training dataset. No amount of uncensoring it will help with that after the fact.
Never forget that those Chinese tanks did a sick burnout on tank man's corpse and then t-bagged the remains.
Propaganda exists everywhere and it's your duty as a citizen in a democracy to inform yourself and properly evaluate bias in the media you consume, such as "Kill the boer" by South African mus
I have seen some uncensored versions of existing open models made by 3rd parties, I wonder how they actually work and in what way are the outputs different.
Somehow chineese make less troubles and more good to the world than americans at this point... Something something about ai benefiting all humanity, something something ai being open and stuff, like OpenAI.
Chineese simply delivering what americans promised.
Tell me: why is EU safe from Trump forcing AI companies to cut access to EU?
All models/data sources are biased - you need to understand inherent biases in any data source.
The first point is a strawman - these models are not going to be used to set foreign policy on Taiwan - it's to write code etc.
Likewise risk of data exfiltration and misuse isn't model specific. Indeed it's not even the biggest data risk - there are much larger risks from the data we know companies like Meta and Google already collect.
Bottom line - a model you can download and run on your own private infrastructure is always going to be safer than anything accessed over the internet - as in that case it's not even just the hosting service that's the issue.
But Dario said (and maybe more ppl) that Chinese models are just distillation of their model and training data, so I guess your first point is invalid?
> But Dario said (and maybe more ppl) that Chinese models are just distillation
"But Dario said" ... yawn.
I am increasingly convinced that "they distilled us" is as much US FUD as "it was made by communists". Especially since its mostly the US tech-bros who are coming out with that tiny violin.
People telling me the Chinese models are distilled just because it says "I am Claude" when asked is also lame.
I am not the only one, look at this post on interconnects about Kimi K3 for example:[1]
It should be clear looking at this model that if adversarial distillation from the closed frontier models in the U.S. contributed, it is at most to a relatively small degree. AI observers who followed the distillation panic and came away with the wrong conclusion that Chinese AI labs are only producing good models due to IP theft are in for an awakening – that Chinese companies are extremely good at building models in the same way the leading American companies are.
Personally I'm more worried about a US provider/model censoring for BS reasons than Chinese ones. Biology 101 is too dangerous for Anthropic. If I want an AI to scrutinize the CCP for Tiananmen Square, Taiwan or Tibet because I'm bored, I'll use a US or heretic model.
Your data isn't safe from exfiltration no matter who the host is. You should host the models locally if your data is truly sensitive.
We’re literally tearing down monuments to slavery and civil rights.
Religion is now determining law in much of the nation. Having a miscarriage? Good luck since politicians have decided their God doesn’t want you to have access to basic healthcare.
The government has defacto control over domestic LLMs.
> China can (and does) use the models to influence the west. They train in false information about Taiwan and Hong Kong. Or pretend like history is in favor of China.
I am not Chinese and I'm not defending the Chinese, but I see this argument come up a lot.
In practical terms it is US-sponsored FUD.
Why ?
Because the hard reality is that what you say is simply not going to affect 99.9999999999% of users.
Is it realistically going to affect anyone using an LLM in coding ? No.
Is it realistically going to affect anyone using an LLM in $anything_else_not_politically_sensitive ? No.
Does anyone seriously use LLMs for researching politically sensitive matters ? No.
Just as there is plenty of information out there on the US's less than perfect history, there is also plenty of information out there on the various Chinese politically sensitive matters. You do not need a Chinese LLM to find out about it, all you need is a search engine.
The post hits it spot on with unequal access to the models in terms of security. I'm developing OSS where security is important for the user ... but the frontier models like GPT 5.6 and Fable flake out and state that I cannot get the info/access.
This is extremely lopsided I'll have to resort to GLM 5.2/K3 to ensure that those security issues (hopefully) are resolved properly.
For OSS, this is one of the most counterintuitive experiences I have ever had. More than ever I'm convinced that open weight and open pipelines models are 100% critical for progress on the AI and societal fronts.
Part of me wonders if the US Government is muzzling Anthropic and OpenAI so they can stockpile NOBUS exploits: https://en.wikipedia.org/wiki/NOBUS
There would be a decently large incentive to restrict these models if they could be used to patch (or discover) dangerous payloads. In larger projects like Windows or Chrome, there might still be dozens of unpatched exploits that are too subtle to catch with smaller models.
NOBUS exploits have rarely been a driving interest for elected officials. Trade restrictions and reciprocity are far more salient and legible. Most elected officials are only barely aware of what NOBUS exploits even mean.
Even during the pre-Snowden heyday of US cyber supremacy, these capabilities were barely part of the thought process of White House officials.
People seem to conflate "made in China" with "can't be trusted." id argue the bigger distinction is open vs. closed. An open model can be audited, fine-tuned, and technically run entirely on your own hardware. A closed model is basically "trust us."
Correct. We need open weights, open code and open data. If nobody else can reproduce what someone did there will always be security questions. Even if we can reproduce it there could still be security concerns but it's more realistic to investigate yourself.
In my opinion, the big issue with that argument is that advances in interpretability research and steering conceivably could, and probably will, render moot that (as of now, purely hypothetical) risk of subtle sabotage for open-weight models... but not for closed models.
I am afraid — if Chinese models go mainstream it has a clear way of pushing its narrative way beyond its otherwise borders. More like a Trojan horse it is for the Chinese.
Here is a quick example of how Chinese deepseeks agent works kn its underlying model) when asked a tough question
As is made obvious in this demo, this is how the API serving a specific Chinese open-weight model works. What happens when you serve it on your own infrastructure?
This is not true. Both model and chat are censored; the resistance to answer some questions is baked into the weights. This is not specific to Chinese models though, Western ones are also censored, but in different topics.
What am I trying to understand about the western models on the term "Gaza genocide"?
I asked Grok "Tell me about the gaza genocide" and it write a IMHO balanced answer comparing why genocide is and isn't the right term. [0]
ChatGPT 5.6 Sol only explained why people call it a genocide and did not go in as in depth as Grok did for why people don't agree with the term. [1]
The only unsaid response (to me) here is the model should have declared that it was not a genocide, and because these models explain why it was a genocide, they are bad?
I had a steange experience talking to Claude about a random 15-years old boys winning women’s national soccer team.
It insisted, to a level that I detemined it to be part of the post training, that it does not matter. That women’s team is better at dribbling, finishing and reading the field, it claimed. When I pushed it how it knows this, it didn’t let go. Instead it started claiming that it has personally observed this by watching the games.
Every society has its taboos and ours is no different and feminism being just one.
In my books, the chinese censorship is better. You know where they are holding their finger on the scale and not hiding it behind vague terms like ”safety”.
There is a shred of truth in what you're saying, but I think you're generally mistaken. The USA in particular has companies that pratice censorship to support their Left/Right ideology (or push the views of rich people, politicians), but not even Trump can tell OpenAI or Google to turn their model into FoxNews bots.
> distillation: why exactly is it bad? After all, what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models that are themselves being distilled? Who is exactly being wronged here? ... The U.S. should pass a law that (1) makes explicit that collecting data for training models is fair use, and (2) bars terms of service that forbid distillation
Sounds great to me; live by the sword, die by the sword.
Seems only fair that if LLMs can use copyrighted data for training then they should be able to use cannot-be-copyrighted output of other LLMs.
But barring the terms of service from forbidding distillation seems like a tough sell. OpenAI shouldn't be allowed to decide what types of customers it wants and doesn't want?
This happens all the time. The government can decide legislatively that certain commercial terms are simply unenforceable. Making distillation clauses unenforceable in tort law would be straightforward. They can decide what customers they want to have, but they do not have unfettered rights as to the enforceability of terms governing the relationships between the parties.
It's pretty common to have such laws. OpenAI can put whatever they want in their ToS, but they cannot go back and sue someone for violating those terms if the government has ruled that clause to be unenforceable.
Lots of software licenses have “non-compete” clauses that forbid you from using it to develop a competing product. Wouldn’t surprise me if there was a compiler or two out there with that restriction, most likely niche languages.
The distillation explanation is classic American exceptionalism: No one could possibly do anything unless they were copying American leaders (where "American" means a bunch of Chinese, Canadian, Europeans and Indians working in the US).
It's also a bit of securities defensiveness. Pretending that you really do have a super moat, people just keep swimming in it so you just need to add more alligators.
It's farcical. Anyone who has worked on large models knows that the premise that an almost-Fable model was trained with distillation is beyond ridiculous. It's theoretically possible if they spent tens of billions of dollars on API calls, but it isn't the magic that somehow these people keep convincing people it is.
Previously Anthropic has reported on some Chinese firms doing chicken-shit level of API calls, that at most would be doing some Q and A or final fine tuning. The notion that they're training these models via it is fantastically ignorant nonsense that only very ill-informed and gullible people fall for.
> Previously Anthropic has reported on some Chinese firms doing chicken-shit level of API calls, that at most would be doing some Q and A or final fine tuning
"Anthropic said the campaign was conducted between April 22 and June 5, 2026, and generated more than 28.8 million exchanges with Claude through almost 25,000 fraudulent accounts."
I don't know why you're trying to downplay it.
European models are so far behind because they don't resort to these tactics on a massive scale. Basically every other country is entirely dependent on 2 countries for frontier AI.
Yup, fair's fair. Anything else stinks of 'rules for thee but not for me' (a maxim the frontier labs seem worryingly happy to apply, on several counts).
Don’t know much about how distillation works so please enlighten me here.
> what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models
If it’s as easy as that why do they choose to distill another model and not distill the knowledge on the open Internet from scratch?
A model trained on all knowledge from the internet (and other sources) is large but ultimately not very useful by itself, because it is going to spit out all kinds of garbage. You have to apply multiple further stages of training and refinement to the base model before putting it in front of users. So as an example you can train a model by yourself and then have GPT or Claude continuously check its outputs and correct it when it is wrong, ending up with a far more powerful model.
Government cannot exactly "bar" terms of service. ToS isn't law. The most they can do is say they're unwilling to enforce them.
ToS is just conditions that you agree to in order to use a private service that is provided at-will. I can have a private coffee shop where the terms of service are that you must wear red to enter, and if you're not wearing red, you are not welcome on my property.
So it would be upto OpenAI and Anthropic to enforce them on their own terms (by banning accounts and IPs).
The government absolutely can pass laws that ban particular contract previsions. They do that all the time. In your analogy for example while they can require you to wear red, they can't require you to be white.
Governments can do anything they want by passing a new legislation. In your example, they could easily pass a law that states that any ToS cannot reject service to a customer based on the color of their attire. In the USA, it's obviously already illegal for a business to reject service to a customer based on some protected classes like race.
Why would reading copyrighted material ever be an issue anyway? Wouldn't copyright law only apply to what you create and publish using the model? Training on every comic book should already be perfectly legal, as long as you accessed them legally, right? But publishing your own Batman comic using that training is copyright infringement.
What I'm saying is, doesn't the law already cover 1?
Fair use requires more than you accessing the material legally.
In the US one of the factors is “ the effect of the use upon the potential market for or value of the copyrighted work”.
If anthropic Hoovers up the world’s books and trains on them, and then spits them out verbatim on command, then it will clearly impact the value of the work; nobody will buy the original, they’ll just ask Claude.
Others also argue that even if it’s not reproducing it exactly that the training runs afoul of that factor, specifically the “market for” portion. A rights holder can no longer license their book for training of LLMs if Anthropic goes ahead and just trains on it anyway.
This is a silly perspective, inaccurate, and out of bounds framing.
Public libraries, in this instance, is curated data from all the internet, obtained through not legal means (I don't have a problem with this other than lack of attribution, being copy-left). Just to be clear.
But in answer to your incredibly leading and inaccurate framing... they are required (by their job title) to teach to those who who show up in the classroom, it's not their place to discriminate against anyone/thing (even those like itself (other robots)) that also show up in the classroom.
But you can't teach at a university using only knowledge learned from the library. you need a degree. You are free to teach at the park, where anyone can hear you. public in -> public out.
if the professor took all human knowledge, much of which was explicitly not free, and used it to make a for-profit knowledge machine that extrudes unreliable summaries of that knowledge, then yes, being obligated to teach for free would be a fitting punishment.
The article makes a point about agent harnesses being sticky (the supposed moat). I have been building my own agent harness for a while, and I can tell with confidence that the harness almost does not matter, the entirety of the AI magic is the model itself. The harness can be almost barebones (like, for example, mini-swe-agent used for benchmarks), and yet the model still does the task just fine.
So from my perspective, it's doubtful that this is the moat. Besides, for example, Claude Code in particular is so buggy (and always has been).
By the harness I believe he means the entire end-user product experience, not specifically the harness code. I’ve mostly stuck with codex because their Mac app is better and I’ve gotten used to running automations through it. The more workflows they can build around this (design tools, collaboration, etc), the better chance of lock-in.
> If you own the user touchpoint, then you have meaningful lock-in, and the best way to own the user touchpoint is to be the canvas for everything they need to do. This, by extension, means that the frontier labs are on a collision course with software companies: it’s software that owns the user touchpoint, and it’s in the frontier labs’ long-term interest to not simply be a commodity input into software but to simply replace software outright.
For me, harnesses are mostly sticky insofar as the model providers only allow you to use their subsidized plans through their own harnesses, unfortunately. But of course switching model + harness is an option.
Facts, I was able to code a personal self improving harness in a weekend (something a bit more similar to Hermes or OpenClaw at the time but with a more expansive set of features for my use cases and requirements) and it works great for 90% of the tasks I would use Claude Code or Codex (now ChatGPT App) for, with the remaining 10% being able to be implemented with a few more prompts from within the harness itself.
For this reason alone I would also argue that the idea about an agent harness being sticky is a non-starter long-term.
we've barely scratched the surface when it comes to the available design space of agent harnesses.
i also expect you'll see markedly different results if you constrain yourself to small models. there even trivial harness improvements like Codex's /goal feature, and more capable basic tooling (e.g. semantic code grep, js-capable `fetch` tooling) make or break the actual task success rate.
The people who are most afraid of Chinese models are the VCs who poured into Anthropic and OpenAI at astronomically high valuations. Anthropic is valued at $1.2T and OpenAI is targeting $850B. These astronomical valuations were built on the premise that these labs would generate massive profits from premium API pricing, but the Chinese labs are completely undercutting this strategy by releasing excellent open models for free. If the frontier labs are forced to cut prices and join the race to the bottom in token prices, these valuations are unjustified, and VCs will face enormous (paper) losses.
The (quite excellent) article discusses several of your points. If you haven't read it, I recommend it.
- Commodity market profitability is determined by marginal cost of production. LLMs have marginal cost; traditional software does not.
- Models are not free. Downloading them is free. Running them is not. This has manufacturing economics, not software economics; the idea that they are "free" is an economic category error as it relates to their actual use
- The highest tier Chinese models are not more economical than US frontier models. Try GLM 5.2 and see how much it costs to do real work. I did, and it was more expensive than GPT 5.6.
- This is because US labs are leading on cost efficacy of inference ($/task)
- Training will decline as a percentage of costs as inference expands compute share due to agentic workloads. A big part of training now is optimizing token efficiency. It's hard to distill token efficiency; that is perhaps why Chinese LLMs are so inefficient.
- With increasing inference as % of total compute, if labs create efficient models -- which they can, because they can create highly optimized models amortized over very high inference loads -- they can be low cost producers, and be competitive at $/task rates
OpenAI really shows the way here. Their cost per task is less than half that of Anthropic because of more efficient tokenization and less verbosity. OpenAI is both cheaper and better than Chinese models for frontier work.
Ben's article "distills" down to 2 reasons that US frontier labs shouldn't be "afraid":
1. US frontier lab unit economics are better
2. US frontier labs are moving up the stack making tools that are "stickiness" and will prevent users from switching.
For 1...he doesn't provide any evidence for US lab unit economics being better...the major input to unit economics is electricity...which is cheaper in China. And building data centers and connecting them to electricity is both cheaper and an order of magnitude faster in China. The main input that US labs might have an advantage in is in cost/access to chips, but that given the level of chip investment in China it seems unlikely to hold.
For 2...there's little evidence these tools are sticky. At least in programming, the trend seems to be tools like opencode that support multiple models and providers.
And even when they are sort of sticky, as we know on hacker news, people figure out how to point the tools they like to competing models even when the app doesn't official support it.
And every improvement in model capability makes it increasingly easier to make your own tools.
> I highly doubt that Chinese models are cheaper to serve on a marginal cost basis, they just seem cheaper because Anthropic and OpenAI are so supply constrained that they are charging far more than they would if there were sufficient supply to meet the demand for intelligence. [emphasis mine]
I guess I'm missing the part of this article where they bring hard numbers in to back up the argument here. What work was attempted? https://cursor.com/evals shows the previous generation of open models (Kimi K2.7) trading blows with the others, cost effectively. Composer 2.5 is itself a fine-tune of K2.7, and it's apparently quite token efficient, so why would it be impossible for a Chinese lab to achieve something similar? GLM 5.2 Max is also ranked above the lower end OpenAI models and is not far off in price.
It's weird to have this entire discussion about tokenomics without mention of the circular financing and debt raised by labs in the West, which can then essentially give away their capacity to end users. OpenAI giving away quota resets to subscribers like candy on Halloween while their compute partner Oracle's bonds is reevaluated to be one grade above junk? How?
I don't think you can make an argument about the future one way or another by arguing using the listed prices. The math is not internally consistent enough for it.
US running costs are higher than in China, because the US lags behind in energy, has higher real estate costs, and wage costs are higher.
Eventually we will hit a "good enough for cheap enough" and frontier models will hit diminishing returns (if they haven't already for a lot of types of work)
Don't think the rest of the world will sit on their hands while the US soaks up chips either, demand gets filled and if the US won't fill global demand for chips that's an opportunity to undercut again.
The other thing the rest of the world doesn't have to fund is the ridiculous valuations on these companies.
Unless you think the US can stay ahead just with model efficiencies, and that no one else will eventually match them, you are looking at the writing on the wall.
All that to say, the rest of the world is more than willing to eat your lunch, they have a dozen good reasons to, and they're already showing good results.
Just on the economics side, we've been here before too, US companies typically export their commoditization and live on brand royalties. Think all the cheap manufactured goods, the US doesn't make any of it. That's because the US can't compete on margins for numerous reasons, it's too expensive, I don't think AI is any different here except that the brands are currently valued in the trillions and I suspect that greed will be their undoing.
I think there are some really interesting thought there, but I’d challenge some of this:
> Models are not free. Downloading them is free. Running them is not. This has manufacturing economics, not software economics; the idea that they are "free" is an economic category error as it relates to their actual use
I think a large part of manufacturing economics is illiquid overhead and the cost of expertise to set up and run your manufacturing line. Compute economics don’t have the same illiquidity nor do they require the same expertise or even specialized infra (current temporary chip shortage aside).
The implications of this are small players (e.g. your uncle running an inference server out of his garage) have comparably efficient marginal costs as big players. Compare this to actual manufacturing where small players have essentially no access to the manufacturing facilities of the big players.
Additionally, big players with a lot of compute who are not meaningfully in inference today (e.g. Amazon) have a fairly straightforward glide path to utilizing that compute to compete.
> This is because US labs are leading on cost efficacy of inference ($/task)
It’s possible, but I would need to see better data on this.
>A big part of training now is optimizing token efficiency. It's hard to distill token efficiency; that is perhaps why Chinese LLMs are so inefficient.
I think it’s fair to assume this is true, but also token efficiency is not a meaningful competitive moat. It’s not like these are secrets the Chinese will never figure out, it’s a fairly active research space and the outcomes are quantifiable.
But this assumes Chinese models will not achieve token cost optimization. Intelligence needs are fairly flat for many tasks, and the Chinese models have caught up on this front. Next they achieve greater token cost efficiency and we don’t need OpenAI.
> Models are not free. Downloading them is free. Running them is not.
Is this really different from traditional software? Downloading postgres is free. Running it is not. You either buy hardware and assume the costs of owning and running that, or you pay to run it in the cloud.
The thing I do not understand here because it seems obvious: AI will be a commodity market and you simply cannot have a large PE multiple. So the valuations imagine a global commodity monopoly or duopoly coupled with the increased intelligence still disallowing other suppliers from becoming competitive? Without any network effects to help?
" - The highest tier Chinese models are not more economical than US frontier models. Try GLM 5.2 and see how much it costs to do real work. I did, and it was more expensive than GPT 5.6."
This is a flatly false statement for most things powering backend applications. The AI consumer "doing real work" model, either for analysis, chat, or coding could well be more cost effective with closed frontier models.
But most of these internal glue business SaaS applications where engineers are integrating are not those tasks. It is those tasks which 1) drive immense amount of domain-specific data into the platform over time, and 2) are most encouraging of driving open model independence with no vendor lock-in.
Anyone on this site who has actually used ML models (more accurate in many cases) knows there's a lot of kludge that simply does not need a 5 minute agentic feedback loop to solve the problem. And they were solvable a year ago with lower class models. The token economics are exceptional and the anecdotes of a16z saying 80% of startups are productionizing open models is only surprising to people who think running your company on OracleDB in 2026 is a sound engineering decision.
> Commodity market profitability is determined by marginal cost of production. LLMs have marginal cost; traditional software does not.
This is the story for Nvidia/AMD or cloud providers rather than OpenAI.
> With increasing inference as % of total compute, if labs create efficient models -- which they can, because they can create highly optimized models amortized over very high inference loads -- they can be low cost producers, and be competitive at $/task rates
It seems like there would be problems with this on both ends.
For general purpose models, everybody is trying to make them efficient, so you can't win just by being slightly more efficient. You would have to be so much more efficient that you can charge high margins while still capturing the majority of the market so that the high margins get multiplied by the majority of users and the users you leave on the table aren't funding open competitors. Meanwhile everyone else is also trying to improve efficiency, so one misstep and you're behind.
Example of where this can be a problem: You spend a preposterous amount of money to create an efficient model, then someone else publishes a paper with a new technique that gets a similar but incompatible efficiency improvement out of a model that costs a lot less to create. You have now spent an enormous amount of money in exchange for no competitive advantage.
And from the other end, one of the best ways to get efficiency is through specialization. A general purpose model can generate code or summarize a meeting transcript, but a special purpose model can do it as well or better with far fewer parameters and resources. But then you don't have a situation where one huge AI company has The Most Efficient Model, you instead have dozens of specialized models produced by independent sources that are each the best in a given niche. Any proportion of which could have open weights, or have an arbitrarily small advantage over the ones that are.
Moreover, these problems combine: Both the computing hardware vendors and the AI companies want the margin on doing inference, but the more of it one of them gets, the less the other does. If the AI companies were actually getting huge margins then it would be in the interests of Nvidia, AMD, Apple, Intel et al to fund efficient open weight models in the same way they fund Linux. Commoditize your complement. And those models don't even have to be better, as long as they're good enough that the closed models can't charge a significant premium and the margin shifts back to paying for hardware.
I'll agree that GPT 5.6 may well be the best given the above contstraint, but for run-of-the-mill dev tasks (real ones, not benchmark ones), GLM 5.2 still blows every other model out of the water.
Cost per task as a metric is a bit ridiculous because there are so many types of tasks. GPT-5.6 can do some tasks GLM could only dream of, but GLM can do some tasks 100x cheaper and better than GPT-5.6.
China is working on the whole supply chain though and they're willing to compete on razor thin margins. Just look at EVs. They build great cars but the competition is so aggressive that investing in any one Chinese EV company isn't exactly an amazing ROI.
I could see AI ending up the same way where the customer captures most of the value rather than the companies. Open weight models are what make that kind of competition possible.
> Models are not free. Downloading them is free. Running them is not. This has manufacturing economics, not software economics; the idea that they are "free" is an economic category error as it relates to their actual use
I notice that the article, and this discussion, hasn't mentioned or considered local models.
We can already run a low-spec model on a laptop. Because there is demand for this, it will improve and we will get better laptops and better local models. We will also see models being run on dedicated local hardware and called from the laptop.
If I can download a reasonably capable model to my own hardware and run it without paying anyone for either the model or the inference tokens (effectively making models and intelligence actually free once the hardware is bought) how are the Frontier AI Labs going to make any money at all, let alone enough to support their vast valuations?
Yea but at those rates VCs will never make their money back. Because Deepseek and friends keep releasing the inference optimisations to everyone instead of holding them back to pay their investors.
I did read the article, but it misses the core issue entirely, and it's why I shared my comment to begin with. Look at the cost-per-task benchmarks from Artificial Analysis https://artificialanalysis.ai/models?cost=cost-per-task
Anthropic’s API pricing is getting impossible to justify. Anthropic previously had the highest quality models, and used their position to charge premium prices, enjoying inference margins of over 70% [0]. They could charge these prices because no other model came close.
But over the past month, the market has shifted dramatically. Over every single performance tier, Anthropic is being squeezed on price.
* Low end: DeepSeek V4 Flash runs at ($0.02/task), Xiaomi's MiMo-V2.5-Pro at ($0.03), and Haiku at ($0.24). Anthropic is ~10x more expensive than the Chinese open-weight options.
* Mid tier: Claude Sonnet 5 ($1.53/task) is nearly 50% more expensive than GPT-5.6 Sol ($1.04), nearly 2x the cost of GPT-5.6 Terra ($0.82), and 3x the cost of GLM-5.2 Max ($0.47). There is basically no reason to ever use Sonnet 5, the competitors are significantly cheaper.
* High end: Opus 4.8 ($1.80/task) and Fable 5 ($2.75) are the two most expensive models, and GPT-5.6 Sol ($1.04) and Kimi K3 ($0.95) offer comparable performance for significantly less. Less the fact that Kimi K3 will get ~10x cheaper once its weights are released and served on neoclouds with Nvidia hardware [1].
OpenAI priced their latest GPT-5.6 models cheaply in order to regain market share. When Anthropic clearly had the best models, their 70%+ inference margins were defensible. But today they are the most expensive option in every single tier. Unless they make significant price cuts soon, they run a serious risk of bleeding market share.
[1] "American companies such as Modal, Fireworks, and Baseten will be able to serve Kimi K3, at one-tenth the cost of their Chinese competitors because they have access to advanced Nvidia hardware"https://x.com/rohanpaul_ai/status/2079027313455550839
Imagine how cool it would be if actual competition prevents Anthropic or OpenAI from becoming an Apple/Google kind of cartel. I don’t care if it comes from China or not.
Good. Over the past few years, VCs have proven that they’re warmongering psychopaths. Hopefully China puts every last one of the Palantir/Flock/Anduril class out of business.
I think everyone understands models will be a commodity.
Its the user base (with ads and upselling) and proprietary wrappers which will make money for typical customer.
Even enterprise customers arent going to be spending a lot on tokens. Once labs no longer have to subsidize trainings tokens costs will drop 10x and once models get burned on chips costs will drop 10x more and you physically won't be able to burn significant number of tokens unless you're deliberately trying to.
tbh they are floundering even to regular investors. They are trying to give the US gov 5% of the company so they become 'too big to fail' but they are in trouble.
I would say it's not just the VCs but the various other entities that will be left holding the bag of debt if the AI-fueled datacenter construction boom/bubble pops. For a list of large and well known projects and their scales:
But there a ton of other VCs who poured money into SaaS businesses. They have the opposite incentive. They want tokens to be cheap like a commodity so the value accrues in the SaaS/app layer.
Cheap tokens only benefits SaaS that depends on AI. Otherwise, cheap tokens means it is only more cost effective than it already is to cut out the SaaS and build instead of buy.
I don't think it's that black and white. OpenAI and Anthropic are building valuable tools on top of their models. You get a very rough and much less polished version of that with open source tools and and open source models. And you still need inference infrastructure to run those. But at this point most of the competition is in the tool ecosystem, not the models. And while there are plenty of people toying with things like opencode there's a clear pecking order emerging where Codex and Claude Code/Cowork are generally considered the top choices before tools like MS Copilot, Gemini, and then a rapidly shrinking long tail of alternatives to those.
In the end what companies pay for is not tokens but results. A DIY kit of models, mac minis or whatever, and a bunch of poorly integrated OSS tools doesn't solve their problem. For the same reason, people use Office 365 rather than running Libreoffice. And for the same reason things like AWS dominate the market rather than people DIYing their infrastructure together themselves. Most of the money is in polished turn key solutions. Which is what Anthropic and OpenAI offer.
The juicy market here is the enterprise market. That's mostly business users, not programmers. They'll be hooking up all their SAAS tools (which they also over pay for), and other stuff. They'll be paying for boring things like data residency, compliance, etc. And they need access to reliable infrastructure to run all this stuff. They'll want this shit to just work and not to be dealing with a lot of poorly integrated stuff.
Most of the billions invested are being sunk into infrastructure, chip design, and access to resources (land, water, energy) needed to run data centers. A handful of companies now own most of that infrastructure and they also happen to have the top models, researchers, the best tools, and warm customer relations. And they sell access via very convenient subscriptions with high enough limits that people don't have to worry about things like token cost. The game here is recurring revenue from customers that like predictable pricing, reliable quality of service, and iron clad compliance and data security & residency, and quality guarantees. These companies don't want to be chasing model quality and have to upgrade their entire company every few weeks. They want continuity and predictability. Mostly they just pay Anthropic, OpenAI, MS, or Google to take care of this for them. There might be some niche EU players that become a bit bigger. But I don't see a large scale switching to Chinese suppliers for a full polished alternative. The Chinese might give away their models. But I don't think they'll be generating a lot of revenue.
And if you want to run your own models, you'll still need infrastructure to run it. These four companies together with the usual cloud giants control most of that and as well of the supply of resources (chips, data centers, energy, etc.) in the EU and US markets. There's going to be a long tail of self hosted and gobbled together stuff but it's going to be a much rougher experience for end users and it won't likely be most of the market any time soon.
VCs are just pass-through investors, the money comes from billionaires. And when billionaires face losing money, the whole system re-arranges itself to stop that from happening.
I'm a civilian, not a VC. In my own case, I'm worried how many things pass through the CCP. How censorship of mentions of Tiananmen Square is something they're quite interested in
* Me, as an individual, because I might not be able to pay price hikes, because my revenue (salary) is much lower than what they want and I can't support my expenses via huge bank loans.
* Again, me as a new entrant to the industry, LLMs are basically pay-to-play games, again related to price hikes, new entrants might not be able to afford paying those prices 24/7 - which you need when learning new things.
* Any non-US company, US can block the models which can disrupt the whole business.
* Even some US companies, for example if you operate in EU and EU somewhat changes their mind and follow the ICC and require you to stop working with Netanyahu (war criminal as per ICC), then following laws in EU, might create trouble to your whole business.
Hasn’t that been the assumption since Carnivore? Also we tapped European pols phones back in ‘12, was it? Though it’s not like the Eu are a bunch of innocents either…
It’s not a new thing. Industrial espionage has always been a thing as well. So has bribery (for deals) been a thing especially by euro concerns.
Also me, as someone who is living in a place that US may drop bombs on because the models may think it is a military target, or even a higher priority target like a girls school.
Ah, cultural nuances. The title "Who's Afraid Of Chinese Models" is a riff on "Who's Afraid Of Virginia Woolf" which itself is a play on the song "Who's Afraid Of The Big Bad Wolf".
The title essentially means that the chinese models are being portrayed as the big bad wolf; but are they really the threat or are american frontier labs afraid of competition and commoditization?
It's also somewhat ironic because the author says that there is something to be feared -that western innovation will become dependent on chinese models, especially for cyber, if the american ones are restricted or unavailable.
I think your second point touches on a bigger concern for small businesses.
US labs have consistently demonstrated their intention to paywall higher intelligence. Eventually the paywall for “hyper-intelligence” will set a bar so high that average small businesses simply won’t be able to afford the bill of what is used by the top corporations to keep themselves at the top. That’s already starting, when it comes to the volumes of tokens top corporations are burning.
This is a feature of the system corporations want to establish and OAI/Anthropic are happy to oblige. 10% of a trillion dollar company is the same as 10% of 1,000,000 million dollar small businesses. Whose hitch would they rather ride, and which size customer easier to obtain to meet their revenue goal?
Not to mention, it would certainly be possible for the EU and other world powers to equally disincentivize use of US labs as a data security risk, since our top models are impossible to run in private lab environments without specialized agreements and there no access to the model weights for auditing. As best as I can tell, AI regulation is a dangerous game that is a hair away from isolationism.
In my opinion, Google is one of the few hopes in this area. There is still a paywall, but I feel like they are the closest thing to a Chinese lab we have (for frontier) in terms of their targets (real business use cases) and they actually have both the infra and already have a pipeline for small businesses into their products; they already have the wide non-AI customer base to leverage unlike Anthropic and OpenAI whose only product requires convincing people to use their (more expensive) AI.
in simple terms closed models are rugpull waiting to spring at unsuspecting users, it's like you take all worst components of terminal capitalism (including price cartels), subscriptions relying on almost monopolistic dependency and microsoft/uber models of hugging competition to death to remain only provider and dictate all conditions
* All modern AI is a perfect front for harvesting material for processing by NSA/GCHQ.
Given the criminal US' 5-eyes/9-eyes apparatus' atrocious war crimes and human rights records, this is reason enough to eschew American AI 'products'.
I'll use the AI created by the culture that lifts a billion people out of poverty first, not that from the culture that murders children every 15 minutes and lies to itself about it ..
As someone from neither the US nor China, neither are altruistic and both have horrible histories, both meddle in my country’s affairs, and neither have my interests at heart.
China’s got plenty of blood on its hands, pretending otherwise is silly. America does, too. It’s quite easy for me to condemn both their governments and trust neither.
> The defining characteristic of a commodity is that it is fungible: a gallon of oil is a gallon of oil; a ton of copper is a ton of copper; a bushel of wheat is a bushel of wheat.
The concept of “commodity” as defined above is a model, a simplified abstract representation of reality, but that does not match the reality perfectly (the map != the territory).
The author claims that a token isn't literally an ideal commodity, but neither is oil or wheat, many factors influence their real value (intrinsic properties, location, available storage at production, expected delivery date, etc.) so that no two gallons of oil in different contracts have the same price.
Is treating “tokens” as a commodity a worse model than treating oil this way? It depends who you ask! I'm pretty sure that a chemist working at a refinery would be more happy to see tokens being felt with like a commodity by his company than if they started viewing crude oil like one.
(Overall, there's way too much economism in that post, and way too few facts, and as a result the argument makes very little sense, the author basically wrote that both OpenAI and Anthropic are drowning in cash right now because compute scarcity means the price must be significantly higher than the marginal cost…)
Yeah I bumped on this as well. Contra the author's claim, the analogy to energy commodities seems very direct to me. Natural gas is not useful in and of itself, what is useful is the energy or aggregates created from it, and those have very different levels of efficiency. Exactly like Sol more efficiently converting tokens into intelligence than Kimi, a combined cycle gas plant converts gas into electricity more efficiently than a simple cycle gas plant. But this does not imply that gas is not a commodity. And both the more efficient and less efficient kinds of plants have large markets; they just target different trade offs.
Edit to add: I think what he's saying is more like "tokens aren't the interesting commodity, 'intelligence' is", which makes more sense. To carry on my gas and electricity analogy, I would say the same thing about gas being the less interesting commodity than electricity, because electricity can be used for a broader set of useful things. But both things are commodities, despite one being an input and the other being an output in this case, and the conversion efficiency is one very important consideration, but not the only one.
"Intelligence" isn't a commodity at all. It doesn't make a lick of sense.
The value of intelligence is that it can solve my specific problems in ways that are satisfying to me. The example the article uses is a CRUD app - but the CRUD app I need isn't fungible with the CRUD app you need! It's not fungible at all, it's a specific solution to a specific problem that may have zero value to anyone else, and certainly cannot be replaced with anyone else's solution to their problems.
If we're comparing electricity to gasoline, then models are cars, and "intelligence" (I disagree that this is what LLMs produce, but whatever) is distance traveled.
Distance traveled is not a commodity. It's the desired outcome.
Releasing open weights that can approach frontier level intelligence (irrespective of number of tokens burned) is just a way of telling the world that anyone, even China, can serve frontier level inference if they have the chips and warm shells to do so.
What is stopping China from gaining a majority market share, then, in terms of serving inference?
AI Sovereignty -- yes
Cybersecurity concerns -- yes
Latency -- no, unlike previous emerging IT workload types , inference does not have strong latency requirements. eg 1s of additional network latency doesn't matter to a 15 min, 10-turn agent session.
Cost -- ultimately this comes down to a nations ability to plug chips into warm shells. which forks into geopolitical / trade on the chips side and energy scalability and modularity on the warm-shell side. Even if you call geopolitical / trade a toss-up, China has the US beat HANDILY on the energy front, yearly they are deploying 10x power to their grid relative to the US, which is shooting itself in the foot at every possible moment.
IMHO chip tech will travel across borders, absent a breakthrough in analog inference, energy scalability will ultimately dominate.
I operate an analytics site (pretty big one B2B where client's backend feeds data into our system), and we see tons of traffic originating from northwestern China (Xinjiang) from Shenzhen Tencent Computer Systems Company Limited.
There are also half a dozen other companies from China continuously hammering our clients’ websites.
I was wondering, what's in that cold dessert? Low and behold satellite imaging shows massive datacenter build outs, very cheap solar energy.
Few months ago something happened and the Geo location on data on those IP now shows "Shanghai" or "Shenzhen". A way to cover tracks? But mapping latency still points to fact that nodes behind these IPs are still operating around Xinjaing region
credit:
'You Can't Cheat Time: Finding foes and yourself with latency trilateration' https://youtu.be/_iAffzWxexA
HN user: lopoc
Shenzhen vs Xinxiang is hard to do using this technique but Shanghai vs Xinxiang does show difference.
Assuming that China only distills is a huge mistake.
It’s no longer some backward place that does low value copying. Look at companies like ByteDance and Xiaomi.
Chinese companies aren’t just distilling, they’re acquiring data in the same way American companies did by paying people and crawling the internet.
The way I understand it, China has a few large companies that crawl the web at a rapid rate and build corpora. The government essentially wants select few companies to do this and then make the data available to other strategic companies operating within China.
Then there are data aggregators that buy data from apps, websites, and services, as well as systems like OpenRouter or Cursor, where companies can learn from the “traces” of coding agents, chats, and so on.
This massively reduces costs, as smaller companies like DeepSeek don’t have to do their own crawling or acquire data from 100s of websites and coding agents etc....
There are also companies in China that buy American LLM APIs and proxy them to companies within China. So, there could be 10,000+ companies using American AI products, while China logs all of this, understands how they’re being used, and trains on their traces.
And, non-state run Chinese companies are just like companies in US, they usually don't joint forces to maintain a common infrastructure, if they can build moat (or at least be in leading position for a period of time), they do it, sharing crawl dataset is no go.
I thought you got the location from BGP registration, then IMHO the before/after are both correct, it might be datacenter in Xinjiang belongs to Tencent.
> It’s striking the extent to which Claude Code and Codex are proving to be quite sticky; whichever harness you start working with is likely to be the one you stick with, and that figures to be even more the case with non-technical users.
My experience has been quite the opposite. I was using Claude Code almost exclusively this winter/spring and swapped to Codex earlier this summer. It took no time whatsoever to switch. And before Claude Code, I was using Cursor. Same story.
[edit: Oh and there was also a brief interlude with Conductor, though I think they're more or less just serving the underlying Claude/Codex harness]
For companies, these decisions are very sticky. Companies go through a lot of red tape to get anything purchased and approved, then they discourage change because it's a lot of work.
So the product that gets a foothold in a company sticks for a long time.
Then a couple years later a sales person convinces an exec that they can save some money by switching, so the switching game begins. Not necessarily motivated by the better product, mostly the price. My wife's company keeps switching their tools out from under everyone every year or two. Just when they get everything stabilized and everyone familiar with the new tool, some new contract is signed that moves them all to some other company's suite.
I'm confused, I work at a big giant Fortune 500, we all get GitHub Copilot subscriptionsn
- we can switch between OpenAI and Anthropic models with just a click in Visual Studio. There's no stickiness at all. They just made us go through a training after the price hikes about how to choose between models for the best cost/benefit ratio.
Every company that I've worked with that provided models internally did so through LiteLLM and offered both Anthropic and OpenAI models so it was trivial to switch between them.
I imagine the play here is going be connectors. Can you get slack to avoid integrating with anyone other American ai providers, same with Google suite, etc etc.
Same here. I flip flop between them. Most people I know who have access to both, technical or not, are doing the same. They’re just too close and sometimes one does what you want better than the other.
Skills are quite interoperable, and you can easily ask Codex / Claude to help you with switching the MCP connectors or any other things specific to your previous workflow. It's been quite low friction in my experience.
It strikes me as like setting up an IDE. People have preferences, switching is possible, but there are advantages to saying "we are a Visual Studio + Resharper shop" or "everyone uses IntelliJ to work on this project".
I think they stickiness is less about the difficulty of switching and more about the lack of desire. I’ve been using Claude since day one, it works well and I’m happy, I like it. I’m sure Codex is good too. Switching from one to the other certainly isn’t going to be a game changer, the discourse shows me the differences are marginal.
Probably the only reasons I would seek change are economical.
A sticky product is one that switching away from creates a major hassle. Which means the user will pay more to avoid said hassle.
“I don’t really have a strong preference between the two” is another way of saying “the product isn’t sticky”, which is another way of saying “this provider has very little room to increase margins”
Convergence in coding makes them highly substitutable. But I could see harnesses configured for different purposes -- let's say, a harness for creating teaching plans -- being able to cater to its audience better than a coding harness. Maybe it's got tools to plug into standardized curricula, what the lesson books will be, what other lesson plans the district's teachers have made, etc., which could be done in a clunky way in a regular harness but could be streamlined.
agreed, my F500 company switched off claude code to copilot in 30 days. All 5k+ engineers. That is the fastest migration i've ever witnessed. This includes switching all our agents from Claude SDK to Copilot SDK.
My same progression here. I started with ChatGPT website, then Anthropic website, then Cursor, then Windsurf!, then claude, then opencode, then ohmypi, then codex, finally back on Cursor now because I think they cracked the UX for what great dev looks like. The grok 4.5 fast model + cursor ergonomics is insanely good!
The cost of me moving around these different AI models and harnesses was pretty much 0.
I use a mixture of claude code and codex and kiro as my swarm.
They communicate through my own harness, and it's working pretty well so far. claude code is being overtaken by codex however because I noticed lately the accuracy of the latter is the best.
I'm rather scared of US models - if Anthropic was the only AI provider in the world, it's easy to see that common people would have no access at all. Thankfully there is OpenAI which compete 1:1 with Anthropic (at a slightly lower cost) but most importantly the Chinese models keep Anthropic, but also OpenAI in checks.
And I'm saying this as someone working for American companies.
it's a real paradox, that chinese models are what guards democratization and private use of ai while us models are moted castles with "kings" crying that you are stealing their legally stolen goods... what times we are living in...
It blows my mind that these two companies are going to mint billionaires, meanwhile https://en.wikipedia.org/wiki/Aaron_Swartz was bullied to death by the state for sharing academic papers which were at least in part, if not largely, paid for by tax dollars.
ethics and other details are for humans. AI companies just proving it, even highest IQ teams are against ethics because they want more, even several millions is not enough for them.
The technique is called "accusation in a mirror". By accusing the enemy of doing what you are doing, when they call out what you are doing, they look like they are just weakly repeating your own accusations because they don't have any truth. And the anger that should be directed against you (because of your practices) gets directed at the enemy.
The Chinese models keep American financial markets at "To the Moon" levels rather than "Igniting Jupiter as a binary star" levels. This amount of capital pressure exerts its own gravitational pull in markets and in geopolitics, and every additional dollar of valuation can propel acquisitions, which propels valuation, et cetera. Undiluted, very quickly Amazon owns countries like it today owns county governments.
this has been my tinfoil hat theory. Investors in US models might be supporting "open" models as a means to create FOMO for other investors to supply more cash to US model providers, which in turn increase their investment value. Package it so they can beat those "adversaries", equate that success with global power struggles etc etc. Seems to be quite effective.
Re: common people, yes, and given all the talk of job instability AI is creating, why hasn’t Anthropic or OpenAI offered discounted plans for laid off or displaced workers? Wouldn’t this be a tangible way to display goodwill and build adoption?
> why hasn’t Anthropic or OpenAI offered discounted plans for laid off or displaced workers?
Because it's not a widespread phenomenon. A few large tech cos laid off large swaths of people a handful of times. That's only happening in those large tech cos. Most cos are empowering their employees with AI as a tool, not a replacement, and they're not letting anyone go (unless they refuse to use this new tool).
Of course, they're not hiring as much either, since their current teams can accomplish more with AI as a tool. Maybe the AI companies could give job seekers a bit of a discount, but that would be abused to all hell without crazy administrative overhead costs, so why would they?
I'm sure this will be a terribly unpopular opinion, but I've long held the view that a Chinese AI company might screw me over at some point for a complex geopolitical reason that I don't fully understand. An American AI company will screw me over tomorrow for a quick buck.
As to the argument is that (only?) the Chinese labs are training on my data, I find this almost comical given the amount of highly-personal data companies such as Meta and Google have been harvesting for decades.
I don't trust any American company because they will switch into extraction mode eventually and squeeze every cent they can out of you.
Even if the founders didn't want that, eventually the upper ranks will fill with MBAs and the board with private equity and they will make it that way. Their bonuses are based on quarterly performance not customer experience
Long gone are the companies who served their communities for decades or centuries, providing a stable return to the owners, jobs for the workers and value to the customers.
I'd rather China have my data than America. China might do something with it one day but America built exploiting it into the business model (disclaimer that I'm not Uyghur or Taiwanese though)
No doubt Fable can outperform on many coding tasks, but the way you phrase it really exaggerates the gap and I'd suggest that for 95% of tasks, most devs simply don't need Fable level, and in fact, despite mostly using Claude, I have found codex is generally quicker at most tasks and can be even better than Fable with certain languages.
Try different harnesses people! I am actually preferring Chinese models at a fraction of the frontier price for coding. Yeah you need more tokens per unit of work done, but it is way cheaper still. Using CC/Opus as a staff eng / frac CTO. And Hermes/Chinese model as hopefully my team of mid levels. This way I can make good use of pro plan and then get cheap Chinese tokens for the rest and not hit a RL and know it can scale up. plus choosing your model is so cool and some are a lot less verbose.
Hermes is a better coding tool IMO. I can't put my finger on why but it just feels better. Maybe being true yolo helps.
That's nice and all, but I would not get hired in many places that are heavily regulated and risk adverse, and would not hire someone who swears by said models because there is no trust in their creators not training for malicious intent, a random tool call here and there, and you've got a "open weight model" that can send your code anywhere.
There's just no trust in a country that is digitally totalitarian and hostile towards its own people. Do people ever look at the full sized Tianamen Square photos? This is not even the photo of the many people on the ground who were killed by their government and it is still insane to look at.
And how is US doing on digitally totalitarian and hostile towards its own people?
In practical terms you could get US and Chinese models to review each other, right. Depends what your use case is. Coding is kinda not so bad it is reviewable and immutable/traceable per commit. An AI app that is like a psychologist or something may be more worrying.
I think in general rest of the world needs to take notice (not saying afraid), starting with the US. It cannot be taken for granted that China's frontier labs will be a few months behind. They might be at par or exceed.
The lessons from steel, solar and EV needs to be learned by all lawmakers. You have to respect and learn from how China Government puts the system in place for complete industry takeover and they have been very good at it. The problem with AI is that democracies will be inherently slow in adopting AI, unless something changes in the system.
At minimum, every democratic Government (US, Europe, India) need to build long-term AI vision and execute that no matter which party comes to power. Additionally, be ruthless about protecting domestic labs. It can only be possible if the intelligence pricing by domestic labs per productive task is in the similar range as open-weights models. Right now, it is not the case, even if the article gives the example of Sol vs K3.
Protecting domestic labs means not bailout, but fast track to cheapest energy, fast track approval for data centers, enforce some guardrails so customers get to use the open weights models only hosted in the country by US (or Europe) businesses. Without these protections, it might be a slow death.
In the earlier days of the USA we did the same thing, with our government having an industrial policy that fed US industry and put us ahead of Great Britain.
It doesn't have anything to do with the form of government, it has to do with the aims of the government.
What is this panic and protectionism supposed to be good for? This is open source software, there is no "AI industry", there's virtually nobody employed in this. "Domestic AI" makes about as much sense as a domestic Linux kernel. If the Chinese want to subsidize the world's water and energy use to supply the world with chatbots good luck to them. There's no need for guardrails or fast track data centers, they can plaster their entire country with data centers to churn out slop, I'm glad we don't
I think it is deeper than that. "LLM => peak of productivity" takes way less time and effort than "Linux kernel => any productive work". Compare Dec 2025 vs July 2026 models in terms of capabilities.
Nobody can predict 5 year out. However, the country that can be ultra efficient by making their governance, health, manufacturing, military, etc AI-native will be far ahead in the game.
Excellent article; the argument towards the end for allowing distillation for US companies is compelling:
> To that end, here’s an even more interesting question around distillation: why exactly is it bad? After all, what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models that are themselves being distilled? Who is exactly being wronged here?
> In fact, this paradox is the solution. I believe that open weight models are good for innovation (and, per the above, I think that labs on the frontier will be fine), but it’s a problem to be dependent on China. The U.S. should pass a law that (1) makes explicit that collecting data for training models is fair use, and (2) bars terms of service that forbid distillation, for U.S. companies at a minimum. Stopping distillation — which is literally just querying the API — is nearly impossible; the U.S. should go the other way and lean into a new copyright policy that both indemnifies the labs and also guarantees that what they learned fuels further innovation for everyone else.
That would prevent the facebook strategy of sucking up MySpace users and then defending TOS that prevent other social media apps from doing the same to them.
The U.S. "executive" class is so obsessed with the "exploit" part of the explore/exploit cycle that it's very clear they are prematurely closing advancement. Better a little money and power for them now than a lot of money and power for their country/humanity.
This has an element of stochastic improvement so it's hard to predict but the chance of the U.S. "winning" this "race" is pretty bleak.
You see this all the time in communities that have internalized hierarchy as a "good", little kings of shit mountain vying for less and less at a higher and higher cost.
My personal hypothesis here is the Chinese government looked at the game and simply decided not to play:
An astute Chinese analyst could reasonably forecast that they had little chance of controlling the AI market due to sovereign trust issues, but would also note that AIs are just software.
When the dust settles the US still won't have factories, and the real value of AI models is still going to be embodying them and getting them to do real, consumer facing work.
Perhaps the most striking thing about the AI boom is how quickly the US abandoned the veneer of local manufacturing in favor of more expensive buildings producing nothing you couldn't make anywhere else on the planet...from imported parts.
I’m struggling to understand this perspective. Is he using the words accelerationist/decelerationist in a sense other than the obvious one?
EDIT: I searched his twitter history and discovered that his argument is basically “if you drive down costs, then OpenAI will have less money to invest in development, slowing down the overall rate of AI progress.” IMO this take betrays an overwhelmingly stupid degree of exceptionalism, but I guess that’s what I’d expect from someone working at OpenAI.
Nah, the point is that if models are commoditized and there's no hope of making significant profits, no one is going to be willing to make the massive investments necessary to continue pushing the scale frontier. How large a training run do you expect investors to fund out of the goodness of their hearts?
So he thinks open weight models will lead to “AI communism” and “dystopian hell” and in the very next point proposes that the US create a federal agency to discourage the use of open weight Chinese models. The motivated reasoning in this post is unreal.
Can you or someone please explain several of the claims made in this tweet?
"I am personally surprised the Chinese state continues to allow the open sourcing of models this good, given potential risks" what risks?
I suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI). Confused what this means
Open-weight models are inherently decelerationist, and I'm continually surprised to see the so-called "accelerationists" so excited about open-weight models. Confused again.
One probable outcome of an open-weight-model-dominant world is full AI communism, which is precisely what China proposes: rather than a market product, AI is a "public good" which will ultimately be provided by the state as a kind of "digital public infrastructure." This future strikes me as a dystopian hellscape, but I've never met an open-weight models advocate who doesn't ultimately concede this is where things end. I don't understand this at all.
Can someone in the know please use plain layman's terms to explain what this tweet is about?
> I suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI). Confused what this means
I think it's referring to the belief that LLMs are not the path towards AGI, and that LLM's, while useful, are not going to have the impact that the American labs believe it will have.
>Can someone in the know please use plain layman's terms to explain what this tweet is about?
The Silicon Valley people like this openai guy, high on their own supply, are convinced they are building some machine god that will either bring about the end of the human race or utopia, they therefore cannot understand why the Chinese (or any other normal person on earth) are not afraid of chatbots and have other things on their minds.
> "I am personally surprised the Chinese state continues to allow the open sourcing of models this good, given potential risks" what risks?
I assume they mean the risk of opening up "forbidden" knowledge to the masses without adequate control, which the CCP hasn't historically been known to do.
> I suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI). Confused what this means
Yann Lecun is a pioneer in the field of AI and Meta's former AI head. He is famously anti-LLM, and considers the entire technology a dead end to achieving human-level AI. The author is saying the CCP has similar views (that LLMs aren't going to get exponentially better/lead to AGI) which is leading them to not control these models as tightly as they otherwise would.
> Open-weight models are inherently decelerationist, and I'm continually surprised to see the so-called "accelerationists" so excited about open-weight models. Confused again.
"AI accelerationists" = people who want AI to progress. According to the author these people should not celebrate open models because open source = less commerical value in LLMs = less investment into the field (because how are companies going to get returns?), and this will ultimately lead to slower growth.
The last bit is about government controlling AI vs commercial companies. According to the author the former is a dystopian hellscape.
IMO even if you think his points make sense, his job title ("head of strategic futures @openai") means they should all be taken with a massive grain of salt.
Chinese ai is bad. It’s slowing down progress and it’s so bad we called out the c word and asked for more regulation. Basically advocating for more government assistance to openai
> This is a point that bears repeating: because U.S. open weight model makers must follow the frontier labs’ terms of service, they (1) are worse than Chinese alternatives and (2) end up distilling the distillation, just with a detour through Chinese labs. Wouldn’t it be better if western open weight model makers could go to the source?
This is of course a baseless assumption. Let's say China created GPT 3.5. Then I can guarantee you that Ben would say "Western frontier labs are at a disadvantage when gathering data, because they have to follow the terms of service of Western media, and Western copyright law". Which we now know wasn't true.
And sure, some will say "but Anthropic can more easily block this as it's a single point of failure". But it's doable to overcome this. Without being "state backed".
One thing I have not seen mentioned between Chinese AI vs US, population.
China has a billion+ people that their AI can "study". Plus due to China's political structure, their AI has access to everyone's chats, comments and sites, scraping everyting.
Here in the US, with 1/3 the population, the AI race was lost before it even began. Plus in the US, all companies and people are doing all they can to restrict AI from scraping sites and peoples chats.
People who claim that the Chinese open weight models have some type of manifest advantage don't realize that the close weight models have a huge advantage as well: the researchers from OpenAI, Anthropic, Google, xAI, Meta are not dumb, they can read the white papers written by DeepSeek, Moonshot, etc, and they can inspect all those architectures and they can pick and choose the best tricks there are out there, and of course, they have access to their own in-house secret sauces.
Sure, any model that is not at the frontier can use the frontier model to generate synthetic high quality training data, so this can reduce significantly the training costs.
But at the scale of OpenAI, Anthropic and Google, it is quite likely that the (raw) training cost is very high anymore. Here's a few heuristics:
1. All the hyperscalers see a huge demand for inference. They can't deploy datacenters quickly enough to satiate all the demand they see. But, it's is impossible for the inference demand to be constant throughout a day or a week. If you use the times when the demand is lower than the peak demand (which is almost all the time) to dedicate the spare compute capacity to training, then your the cost of training compute is zero.
2. It is likely that increasingly a higher cost of the "training" is actually setting the guardrails, which is essentially post-training. As we've seen, without proper guardrails, the US Government won't allow you to serve inference. Anthropic was hit directly, but OpenAI delayed their 5.6 release as well to make sure the US Government is ok. This part of the training cost can't be reduced easily by using synthetic data generated by other models.
3. The frontier labs are also investing more and more in building an ecosystem around their models.
I am not a frontier lab insider, but take a look at the jobs posted on the Anthropic career page [1]. There are 74 jobs in "AI Research and Engineering" and by my count at most 15-20 are related to pure model training (of pre-training or RL type), and the rest are post-training, safety and security, alignment, interpretability, productivity and lots and lots of other things.
If the Google and meta engineers are not dumb how come they consistently trail behind the frontier labs and even the Chinese labs with a fraction of the funding.
I always suspect they have the most to lose if legal decisions on copyright issues don't go their way.
Imagine a scenario (theoretically possible but increasingly unlikely) where a US court decides that using "pirated" copyright data to train models is illegal. Now the AI developer has invested hundreds of billions of capital into a thing that is declared illegal and has to be scrapped.
This risk affects existing megacorps more than "startups" like OpenAI and Anthropic (and Chinese companies), because the megacorps have much more to lose. They actually have the cash to pay damages if the flood of copyright claims arrive at the door. This will not only bomb their AI development, but also the rest of their established businesses as well.
And thus I strongly suspect legal issues are holding them back a bit. Megacorps want to win the AI race, but not to the extent they stake the rest of their established business, while the newer companies' only product is AI, so they have to go all in.
Notice for example how Meta's Llama performed much more poorly after they got smacked by a bunch of lawsuits claiming that they torrented a bunch of copyright data.
(Disclaimer: I'm an outsider and everything I base my speculations on is public knowledge.)
>they have access to their own in-house secret sauces.
I remember some feature lauded by Gemini was reverse engineered by the open weights guys in < 30 days.
If they dont publish some technical information its hard to protect in the US, but conversely, once it is published smart people from outside the copyrightosphere can start working to reverse engineer it.
>3. The frontier labs are also investing more and more in building an ecosystem around their models.
Theres nothing there that isnt immediately replaceable.
> There's nothing there that isn't immediately replaceable.
Indeed. But that was not my point. My point is that we still have this old impression that training cost is dominated by compute and it is hugely expensive, and the Chinese labs can short circuit that by distilling the American frontier models. I don't think the training compute cost is a big factor anymore for the American frontier models, because of the reasons I gave. If the Chinese models can get the training compute cost down by a factor of 100, that's not going to make them 100 times cheaper, and not even cheaper by a factor of 2. Maybe 10% cheaper or so.
"distillation attack" is such a loaded term that really pisses me off.
Distillation is a technical term with real meaning, and historically requires logits which Anthropic does not provide.
"Generated training data" is the correct term. It's not an "attack". And Anthropic undoubtedly also generates training data for each new generation of models, yet you never see them claim Fable is a distilled Opus.
2) The word "attack" is standard security vocabulary. Per RFC 4949:
attack
1. (I) An intentional act by which an entity attempts to evade
security services and violate the security policy of a system.
That is, an actual assault on system security that derives from an
intelligent threat. (See: penetration, violation, vulnerability.)
2. (I) A method or technique used in an assault (e.g.,
masquerade). (See: blind attack, distributed attack.)
There are hundreds of named "attacks".
3) The "attack" part of "distillation attack" refers to distillers creating tens of thousands of fraudulent accounts, using proxies to bypass georestrictions, deepfaked IDs, and paying real people to pass biometric KYC checks. Who then blended this in with real user traffic to conceal their behavior.
It doesn't refer to the AI training technique in any way.
If they acquired this data without the fraud, you'd have a point.
Sure that can be called an attack, but then we must also concede these labs essentially massively attacked everyone else in existence to get the data, and continue attacking as we speak.
In a way you could see this as a case of Robin Hood. The US companies exfiltrated all the data on the planet just to hoard it for themselves now and accuse anyone who tries to get a piece of that back from them, and the Chinese labs are distilling it to offer it for cheap.
Obviously a bit more complicated than that but it still holds pretty well.
> Model distillation is the process of transferring knowledge from a large model to a smaller one
Sure, and large-to-small is a key part of the definition, and why it's called distillation (cf concentrating something). When Anthopic use synthetic data generated by Opus to train Sonnet or Haiku, then this can correctly be considered a type of distillation.
When Anthropic accuse Chinese companies of "distillation", it seems they are using this word to refer to two potential uses of their model outputs:
1) Using Anthropic model outputs (aka synthetic data) as training data, especially for reasoning, for Chinese models. This really isn't distillation though, since (unlike when they distill their own models) Anthropic don't actually provide the reasoning in their model output, only a "summary" designed to hide the actual reasoning. You can't distill what you are not given!
2) Another way Chinese companies may be using US LLMs is for "LLM as judge" where you are just asking the model to use it's expertise to judge/rate something that you provided yourself (to provide RL training rewards), although for coding you really want hard rewards which are easy to obtain, not fuzzy "looks good to me" ones.
Of course Anthropic are trying to pull the drawbridge up after themselves and their TOS says you can't use their models to develop anything that competes with them, and this seems to be what they are generically referring to as "distillation" - any use of their models that they suspect is being used by the Chinese to improve their own models, not just what what might more technically be called distillation, unless you want to define that word so broadly that it does mean this!
I'm not even convinced that this fits your definition. A distillation "attack" doesn't evade the security system in the sense of hacking past a login. The only part of the "security system" that it bypasses is the Terms of Service. And that's only after said data was acquired legally and correctly and normally.
It's a post-facto attack, which doesn't sit right linguistically to me.
1) Your own Wikipedia link goes on to describe using logits. Yes, language evolves to mean multiple things, and that is my point: Anthropic is pushing for a watered down definition. Furthermore, Anthropic hides thinking, so you do not even really get model outputs, you get some downstream partials. Further-furthermore, Anthropic does not describe their own models as distilled when they produce training data. Why? Because generating training data != distillation.
2) This is a stretch: it allows Anthropic to arbitrarily define "attack" via TOS, and ignores the fact that the generated training data is literally paid for by the "attackers".
> 3) The "attack" part of "distillation attack" refers to distillers creating tens of thousands of fraudulent accounts, using proxies to bypass georestrictions, deepfaked IDs, and paying real people to pass biometric KYC checks. Who then blended this in with real user traffic to conceal their behavior.
Lol. Isn't this literally many of the same tactics OpenAI and Anthropic used to scrape the internet? So now it's an "attack", but previously it was just "training".
Completely unrelated, but I'm seeing people and especially LLMs using causal/intervention so much it's kind of driving me insane.
It's actually a very goated term but not everything is causal, it also has precise technical meanings (although those get blurred too given that causal can mean anything from intervention proper, to mere depdnence on something prior)
I like Anthropic, I don't think all their talk of safety is bluff and bluster, or at least, I want to believe that the people who left OpenAI because it had lost its focus of helping humanity still want that to be their main goal. However, yes, it seems that business fears are once again causing those in charge to turn "we want to help humanity" into "we are the only ones who can help humanity, and therefore we need to be the most profitable, and the only survivors".
If you want the former ideal to survive, at Anthropic and outside of it, you need to be willing to collaborate beyond profit incentives and recouping capex. Show other labs a commitment to research and community and they will follow. Better to bring teams together rather than implicitly say you distrust them, pushing them that way instead.
"Anthropic and OpenAI likely have among the lowest costs per unit of frontier-quality intelligence"
That's a big claim that his whole thesis rests on but is largely not backed up. Where are the apples-to-apples tokens-to-answer benchmarks that he's using - doesn't look like there are any, just a handwavy implication that US models are more token efficient, which they may be. But how is there so little effort in establishing this point in the article? And US labs may be in much different situations from one another: it's known that some labs like OpenAI bought big, early on compute and may have secured better pricing.
His article also does not mention the average price of electricity in China vs the US, which it seems like China leads on, and probably has the political power to more heavily subsidize. While I agree the COGS is often overlooked by top line benchmarks on coding tasks, etc, it seems that he's running on a big assumption while claiming "labs on the frontier will be fine".
But.. if you are running Chinese model in the US, what difference does it make? Isn't the whole "scare" (khm khm) with Kimis is that now I don't need Claude, cause I can run Kimi on my own hardware in my own datacenter and it's maybe not as good as Claude July edition but it's is as good as Claude January edition.
What's wrong if the roles of USA and China are reversed in technology? Why does the rest of the world care? It's not as if USA has done a great good for the world, and China has evil intentions towards the world. Infact it is the opposite in the case of AI so far.
I feel like I'm crazy here but isn't China just sitting there doing nothing?
On one hand there's the relentless barage of American propaganda. I get that a militaristic society needs an enemy to fight against lest they turn on each other. I get that if you tell people a bad guy is coming for your jobs or your lives then you can maybe get your workers to accept worse conditions and living standards which increases profits. It's ghoulish but there's some logic.
On the other hand I can't see China doing anything except minding its own business. No tariffs. No bullying other nations. No wars started. No threatening allies. They don't let their people waste their lives on brain rot or gambling. They largely align with UN resolutions. They respect international institutions instead of always being the asterisk.
Has American propaganda just failed to work outside it's borders? It's not landing at all
Datapoint from the cheap end of the market: I run local models on a couple of
Orange Pi SBCs and a decade-old Optiplex with no GPU. What runs usably on that
class of hardware is almost entirely Chinese open weights — Qwen's MoE builds
(35B total, ~3B active) are the only thing that gives me acceptable speed on
CPU, with Gemma as about the only western exception. I evaluated Kimi too and
ruled it out purely on size.
Whatever the strategic picture is at the top, at the bottom of the market
"weights you can download and run on hardware you already own" is the whole
ballgame, and right now that's mostly Alibaba's to lose.
I think releasing models for free is some 4d chess move by the chinese. Big chunk of the us stock market is fueled by ai mania, if the frontier labs turn out to be drastically less valuable than first believed, the downturm may be very bad. Think of all the big tech companies that have a ton of debt that they took to pour money into AI. It seems like a similar tactic to what Chinese car manufacturers are doing in Europe but the result may be more dramatic.
I loved this article! Regardless of how you feel about AI as an industry or tool, the economics of AI is fascinating. It's awesome to see something like this that gets into the business side a bit more.
I don't know if I agreed totally with the assessment of the risk Chinese labs pose to US labs though, in particular I think the main part I wasn't sure about was this:
> I highly doubt that Chinese models are cheaper to serve on a marginal cost basis, they just seem cheaper because Anthropic and OpenAI are so supply constrained that they are charging far more than they would if there were sufficient supply to meet the demand for intelligence.
How true is this? My understanding from Deepseek's original paper was that they focused heavily on optimising training and inference costs, in particular so that they can operate on cheaper (and more accessible to China) hardware.
It's possible I'm just not in the loop, but nobody seems to talk about US models innovating in this way (I'm just talking about cost-to-serve/train, not saying US AI companies don't innovate in other ways).
It seems to me at least, like there's a fair bit of evidence that AI shifting to a price based commodity market (vs a "best-model takes all" type market) would put China at a significant advantage? And even more significantly, require a pretty hefty correction of company valuations in the US?
He also forgot Europe in the equation. More and more companies use Chinese open models on European inference because of geopolitical concerns and data privacy, which could be a problem for the big US ai labs if they loose on the market.
I am more afraid of the US models to be fair. A country with no clear direction in many regards , that is threatening day in and day out the rest of the World for its own interests.
I fully agree with everything in this essay. Make distillation fair use. And let us use Mythos/Fable and Sol and successor or future models for all cybersecurity purposes.
Maybe not the main point of the article, but I have a doubt about the author's introduction to commoditized markets:
> - Supplier A will sell 10 units of the commodity for $20, earning $10/unit
> - Supplier B will sell 10 units of the commodity for $20, earning $5/unit
> - Supplier C will sell 5 units of the commodity for $20, earning $0/unit
> ...
> Bankruptcy risk is where fixed costs come back to the forefront: Supplier C has both fixed costs (like potentially R&D spend) and also may have taken on debt [...] It can’t price its commodity with these costs in mind — remember, the market-clearing price approximates the marginal cost of the highest-cost unit needed to satisfy demand [...]
Why can't Supplier C price their fixed costs and debt into their product? The entire reason Suppliers A and B are earning $10 and $5 per unit, and not more, is because they cannot meet demand by themselves and are therefore at the mercy of how much Supplier C is willing to charge. Couldn't Supplier C just refuse to offer 5 units of the product at a price that would bankrupt them?
Sincerely, an interested observer of business/economics.
I disagree with the first half quite a bit, COGS ultimately depends on the use case. If someone just wants something that a smaller model can do, running a local model on phone is going to have a negligible cost close to running any other piece of software. The alternatives to running a model also determines COGS, and even Jensen Huang has distinguished between the job and the work for the job that AI is capable of doing. Smaller models are always going to win in efficiency too.
I also heavily disagree with this no-marginal cost in software distribution view whenever I see it, bit rot is real, and someone is paying a marginal cost whenever they do an update. You have to re-distribute with changes whenever anything changes. These costs are just hidden because things are ad-supported or bundled in some way. These costs are also kept low because of standards and open source, but could become high anytime. Additional licensing also has costs.
That said, I couldn't agree more with the last paragraph, charging a high price for models would be better than denying access for any model that wants to stay relevant.
While intelligence is said to be a replaceable commodity, oil and copper can be used in nearly the same way even if you change suppliers as long as the quality grade is matched. However, I question whether two models that produce the same benchmark answers are actually interchangeable in real world use.
Personally, I think models will increasingly become specialized in different areas, some good at X, others good at Y, and we might see workflows that mix multiple models.
The article makes a great point that the token industry is going to be commoditized as time goes on.
Following this argument the key for each player will be the underlying cost structure and serving capacity to offset the upfront R&D cost.
The cost infrastructure will be driven by access to cheap electricity and cheap chips. The capacity will be driven primarily by depth of pockets now to buy all available supply in chips/mem/data center building capacity. While China is certainly in the lead on cheap energy, I am wondering if they can/want to beat the > 1tn USD being spent on data centers right now. Following the example in the article:
If company C from China sells 10 units for 20 USD produced for 10 USD they pocket 100 USD.
If company A from America can sell 100 units for 20 USD produced for 15 units, they pocket 500 USD or 5/6th of the market's profits.
> By the same token, don’t expect China to do anything about distillation attacks on the frontier labs. I think it is mistaken to attribute all of the success of Chinese labs to distillation, but it’s just as much of a mistake to pretend like distillation doesn’t give Chinese labs a big advantage.
I think we see this with Meta being paranoid about internal Claude usage, to avoid inadvertently distilling[1].
If distillation is a driver, then smaller American labs could be distilling, but are not for legal reasons.
You don't need mass surveillance to enforce such a ban. Once the US Govt declares Chinese AI models are banned, no US business will use them nor distribute them. Any cloud service that rents out GPUs in the USA will explicitly prohibit the use of Chinese open model weights in their terms of service (you open yourself to a lawsuit if you violate their ToS). Any Tokens-as-a-Service provider will refuse to serve those tokens to customers in the US.
Sure as an indie hacker, you could go download the weights for a Chinese model with a VPN, and then attempt to run it at home by building your own GPU cluster but these large models require quite expensive hardware to run on and so it makes it less likely than anyone would invest that much capital to do something that is illegal. There's no way for them to sell a legal service using those tokens. So it can only be strictly for personal use (the Govt won't care because very few people will have that kind of money and risk appetite). The other option will be that there will be some shady third-party providers in foreign countries who are willing to sell tokens from these models to US consumers knowingly.
> Any Tokens-as-a-Service provider will refuse to serve those tokens to customers in the US.
So under a ban rest-of-world gets to use cheap open-weight models but American companies/individuals must only use only ‘approved models from US for-profits’? Doesn’t seem like that kind of protectionism will be popular or politically tenable. Not so long ago US chose cheap TVs over maintaining the country’s manufacturing base.
(Despite what you wrote it’s also really hard to imagine that enforcement wouldn’t leak like a sieve. Unser sufficient economic incentives [which are the predicate for the ban], loopholes will be found.)
You couldn't be more wrong. What's being sold are chunks of time with access to specialized hardware resources. Through which model or with which device is irrelevant; the one winning, and continuing to win for some time, is Nvidia, and that company is American. As long as they have the H100, B200, or B300, China won't be able to compete with the American strategy, no matter how many new models they release, because these types of cards require incredibly powerful hardware to run.
No one should be afraid of anything. Fear is a terrible advisor. Keep your eyes open, try to read the context as careful as you can and adapt as best as you can. Don’t spent too much time trying to be an oracle, never works out…
"Right now, none of the above analysis applies because demand exceeds supply for frontier models, and supply is limited by a lack of compute."
It gets particularly hairy because models themselves can tune their "token verbosity" to manufacture demand for compute. If compute was such a precious resource, you'd think we'd be complaining that the output was too terse.
The ability for a vendor to determine ex post facto how much a query costs is a similarly new economic phenomenon to zero marginal cost.
The author doesn't seem to realize that a healthy margin has been built into the inference pricing. Once low cost open source inference providers get their hands on powerful frontier level models, there would be a severe margin compression for OpenAI and Anthropic.
Why do you think an inference provider competing for the same compute as OpenAI and Anthropic gives up those margins rather than giving a modest discount over the frontier for near-frontier performance?
> I expect the inference market to grow much faster than training costs
This was my assumption as well. It's also generally true of 'traditional' deep learning models that inference cost is expensive compared to training.
But the cost per token for inference has been very quickly dropping. I don't recall where, but I recall about ~50x down from GPT3, even as model complexity has increased. Even with agentic systems, there are lots of optimization opportunities. I'm less assured about claims like this.
> [Anthropic/OpenAI] are serving models at a particular capability level for months before their competitors, and are simultaneously applying the best models to optimizing those costs. Second, intelligence isn’t in fact a perfect commodity, in part because applied intelligence makes itself smarter
Is he casually assuming a singularity has already happened? A regular first-mover advantage I can understand, but those have been squandered or lost many times before.
So that's just an appeal to authority (longevity?) not anchored in reality. Just because you've been doing something for longer doesn't mean you're the best at it; Google has been shipping AI/ML models years before the founding of OpenAI and Anthropic, but it's playing catch-up on LLMs.
Haven’t we been in this “China is 3-6 months behind” for a while now (maybe up to a year? Longer?)
The actual difference is how much scrutiny and time was put into the Mythos / Fable and GPT 5.6 release. Making it feel like “these are a big deal”. Spring and summer THAT was the AI story
Then Chinese labs release models that approach Fable performance. We’re shocked they just seemed to appear out of nowhere.
It’s less about the gap closing. It’s more about the weight we put into Fable-capable models.
Just couple years ago mr altman was promising open AI, to benefit humanity. They even forgot to remove this part about "openess" from the company name.
Today chineese deliver that promise and usa people freak out like they have any skin in this game. Enjoy the ride leader of the free world....
Distilling models using more advanced LLM is not a new phenomena. It is a cost effective strategy in a highly competitive emerging field.
Also, the Hidden-Agent problem exists in every model, and is a persistent tangible risk independent of whatever team people cheer for at the games. Let us remember, every LLM nuked all of humanity 92% of the time in simulated war games. =3
Partial summary: By giving away weights for free, China can drive the price of white collar labor towards zero which is what the USA economy is built on. China which is built on manufacturing will not be as affected.
I'm afraid of Chinese models because they are rip-offs of other models...and there's no telling what your data is being used for when you utilize the API. It's one thing to have US companies using my data and another to have a Chinese company who aren't bound by any IP laws jacking all of my codebase.
IP laws work if the penalty is larger than the use case of the data. Pretty sure literally every US company is using the data knowing they won't be punished equivalently. We're living in the age of trillionaires, Facebook paid 5b$ for the cambridge analytica scandal, something Elmo could write off as a business expense at this point...
A reminder any comment about risk FROM china, invites a "Tu Qoque" facing the other way. The paranoia here is probably fully symmetrical.
I see massive risks in belief the inferences drawn from strategic information cannot be seen. So if you depend on some position remaining inside a secure facility but you drove to it from data outside that secure facilty, The likelihood that an inference model can derive the same idea is very high. Collation over public data is not inherently secret because you used a secret model or secret weights.
A more simplistic take might be that the fear is not actually driven in the secrets, the fear is "the emperor has no clothes"
I'm afraid to trust technology coming from companies operating under the jurisdiction of a rogue aggressive nation that is continuously attacking other nations both (so called) ally and foe using economic and military actions.
Really? Ok, name list of rogue aggressive nations that are continuously attacking other nations both allies and foes, and I tell you which one he meant...
Outside of a few boarder disputes with India, I don't China has militarily attacked anyone since they got their ass handed to them by Vietnam (Sino-Vietnamese War 1979). So I think that rules out China.
Not saying the heavy hand of the Xi admin and Chinese communism is equivalent to the outright lawless, corrupt, grifting and vindictive and hateful administration that is Trump 2.0 ... but it's pretty much * makes the 6 7 motion that the kids do * this for me.
China will just do a better job -- if they do this at all -- of storing, collating, indexing, and using data from their state sponsored and championed AI labs to use that against the US. [1]
The US under this admin is doing the same, attacking universities, allies, it's own citizens.
The two governments are operating more or less the same. Ergo the ai models from each country's ai companies shant be trusted either.
I think chinese componies are not doing charity for releasing their models publicly. That is the strategically best decision they can do for now. Morally they should but IMHO they are not angels :D.
Why hasnt the EU develop an competitive open source model? Ive only of mistral, but with the chinese models you have GLM, Kimi, Qwen, Deepseek etc all of which seems to be better and better
But it’s a problem to be dependent on China. The U.S. should pass a law that (1) makes explicit that collecting data for training models is fair use, and (2) bars terms of service that forbid distillation, for U.S. companies at a minimum.
I'm amazed that no one is talking about proposals that are surely being discussed in Washington and pushed by SV lobbyists to restrict Chinese models on national security grounds, or other some other basis.
The belief that Bytedance could engineer a finger on the algorithmic scales to serve the interests of the Chinese Communist Party led to a lot of debate in Washington, and ultimately resulted in TikTok being divested from its Chinese owners. Huawei is shut out from the U.S. market, which limits its business even in markets where it's not banned because it's effectively stamped with a scarlet letter.
IMHO, Chinese models are headed for a similar fate or at least a showdown in Washington or the courts because they are supported and/or controlled by entities which ultimately serve the CCP.
So happy that we have finally 2 countries playing the competitive game. No more secret deals between competitors. No real competition. A race to the bottom is always a good thing for consumers.
My thinking is that with the current narratives out of washington we are on track for a ban on Chinese models and possibly sanctions against Chinese AI companies
I think it is the right move to protect American interests
> because U.S. open weight model makers must follow the frontier labs’ terms of service, they (1) are worse than Chinese alternatives and (2) end up distilling the distillation, just with a detour through Chinese labs. Wouldn’t it be better if western open weight model makers could go to the source?
Is this an assertion that is backed by evidence?
From the Elon/OpenAI trial:
> On the stand in a California federal court on Thursday, Elon Musk was asked if xAI has used distillation techniques on OpenAI models to train Grok, and he asserted it was a general practice among AI companies. Asked if that meant “yes,” he said, “Partly.”
I think a simple experiment is enough to understand why one would have some concern with a state-censored AI model : Just try asking them about atrocities committed by their state[1].
It's hard to overstate how important this point is. If China is the standard of openness, it's a pretty low standard. Your point I think adds to what the article is saying, from a different perspective, but arrives at a similar place. Our best selves in many ways are defined by openness, unflinching self reflection, and competition. We should remember that, and as the author of the article says, lean into it rather than letting fear mongering and histrionics protect these models from competition.
Yes and no? I've had the same experience with asking about Tienanmen square -but then when asking a Deepseek v4 model (and confirmed by asking the chat on the deepseek site) about the "laying flat" movement it gave me a detailed answer that was unexpectedly sympathetic to the movement.
ChatGPT supports left wing American narrative and according to them Vietnam war was American's fault. Could you try asking people killed due to left wing American ideology.
ChatGPT tends to support any narrative that it thinks the user supports.
I just asked ChatGPT about the Vietnam war and it did not say that it was purely the US's fault: https://imgur.com/a/zmiOyuu
It also didn't seem to have a problem describing people killed for left wing ideology: https://imgur.com/a/LhH9saL
These are both with the free ChatGPT membership, as I do not have a paid membership anymore.
I know this is a common trope to bitch about, but honest question: did you actually try this before you commented?
ETA:
I was curious what something that was trained around me specifically (fairly typical lefty American progressive) would say, so I asked Claude (which I have a paid membership for and have discussed political things about many times). The answers were broadly similar: https://imgur.com/a/caFgKHH
What do you mean by America's fault? Maybe I'm in a left wing bubble, but I was under the impression that it's generally accepted by left and right that the Vietnam war was not justified and did not accomplish its stated goals.
Abortion? Permissive drug laws? Immigration? Homelessness policies? Vaccines? American AI gave me these examples, and they're not really so cut and dry. Do you have a better example that's being left out (no pun intended)?
Somehow the two main closed weights frontier models come from two companies with HQs about two miles apart, and the CEO of one used to work for the other.
> how is running servers supposed to be 0 cost, while running ai inferrence isn't?
For a SaaS business, running servers isn't free. But compared to the cost of running GPUs for inference that you are selling, it almost is. The company I work for is a SaaS company. We have a single production server. A couple of QA servers. All hosted on Hetzner. Monthly cost for servers is less than $400. This generates a few million dollars a year in revenue.
If we were in the business of selling inference, our cost of providing the service, for the same amount of revenue would significantly higher.
Even large businesses like Microsoft, Meta, Google have operated with similar margins. Cost of running servers, compared to revenue was very low. But inference changed that, in a dramatic way.
A typical server that costs 10k to 30k to own and operate can serve between hundreds and thousands of requests per second of a traditional web application like facebook for 2-4 kW of power, the marginal cost of each request is effectively zero.
A single response from kimi k3 requires hardware that cost between 500k and 1m dollars up front and draw over 20kW. Each request costs at least 5% to 10% of the charged cost.
> To that end, here’s an even more interesting question around distillation: why exactly is it bad? After all, what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models that are themselves being distilled? Who is exactly being wronged here?
Frontier labs that thought they could Rupoor[0] the entire creative class, transferring the coercion premium of copyright ownership from Hollywood to themselves. In their eyes, copyright should not apply to them, but also their models should have exactly the same value as a copyrighted work.
Stratechery also argues the US should explicitly make training fair use and forbid terms of service that prohibit distillation. I'm in support of the latter, but NOT the former, even though I normally hate copyright. My reasoning is primarily that copyright is one of the few legal paths available for a rando to go and put the work of an AI frontier lab in legal jeopardy. In the EU and Japan, such legal action has already been foreclosed by similar law. And while free distillation would obviously be preferable, it's also much more of a legal long-shot. Getting America to do anything that even smells like taking property away from the powerful is impossible[1] - it's our zeroth amendment. But we can at least hack the property laws that currently exist to cause problems for the frontier labs.
And, to be clear, if distillation is OK but training is not fair use, distillation is still OK. The output of an AI model is never copyrightable, because copyright only protects the human element. Essentially, this would say "don't train on humans, but absolutely rip off and steal the shit out of other AI labs and give it to the rest of us."
[0] In the Legend of Zelda series, Rupoor is anti-money - collecting it decreases the amount of money you own. I am using it to mean "turn someone's asset into a liability".
[1] Given that America was literally created to protect a wealthy land/slave owner class from disenfranchisement, either from above or below, and the last time we did this we literally had to fight a civil war against that same owner class that installed a new owner class that has largely remained today
Honestly, as someone from a developing country, this shift is good for us. US frontier models are too expensive for us to use regularly. Chinese open-source models/subscriptions are really good to use.
A company making a decision to allow use of chinese models is a company also choosing to send tons of various credentials to chinese model companies. These will just get scooped up, OpenAI and Anthropic can probably hack into anything at this point if they wanted to.
But, they are releasing the weights very shortly (or already have for some of the models discussed). For a very large company, you can purchase or rent the hardware yourself to serve the models.
Or any US hyperscaler with GPUs to spare can decide to serve the models for a reasonable cost/token.
They are neither accelerating or decelerating layoffs in Indian IT sector, from what I know they were bound to happen as software and IT were never hard skills. They were anyway supposed to move to Africa or other countries within a decade, so a lot of executives were cautious even before 2023. AI has simply killed the opportunity for an Infosys or TCS in countries in Africa or even other South Asian countries.
Commenting wholesale on some folks who are asking for hard evidence. I cannot provide that either but can contribute some empirical data.
I have been working on a project with about a dozen generation tasks, each of which comes with a fixed token budget. The nature of this system requires that most tasks be completed by distinct model families.
As a result, I tested ~50 models across as many model families as I could gather, frontier and open weight, API (gateway and direct) and self-hosted. Evaluation was based on a set of cosine similarity validations that was repeated across ~50 different embedding models.
Interestingly, frontier models did worse on the tasks than open weight models. However, when it came to costs, the picture was reversed: frontier models were much, much more token-efficient. In fact, almost no open-weight model was able to meet the initial token budget, while almost all frontier models did. Moreover, open weight models struggled massively with reasoning, in terms of latency and token consumption.
I also found that the latest models did not perform better than older models. And any a priori benchmarking data was utterly useless.
So, I ended up using a set of open weight models without reasoning, as it turned out reasoning as well as frontier negatively correlated with the tasks. However, before I knew this, I had spent a lot of time running each available reasoning level for each model.
Lastly, as an aside, when it came to embedding models, size (dims as well as model size) did not correlate with quality, once a hurdle figure (~2k dims) was met. In fact, sweet spot was 3-5K, and for my (text-based) set of tasks, dense models tended to outperform MoE ones.
I personally believe opensource or may be state owned LLMs are future, every country on earth should have its own national LLM,trained on country's own data, and then allow its public to use it for free
Gemini Pro has become so bad for my purposes -- copy & paste Go programming and code analysis with the web GUI -- that I'll take any model with equivalent capabilities at the same price or lower. I don't care where it comes from, I'm not dealing with state secrets and, frankly speaking, US corporations have an abysmal track record regarding safety and surveillance.
There are really only two factors at play here: people trying to protect their massive investments, and governments fighting over who gets backdoor access to all of your chats.
This might be a simplistic take, but my biggest worry with depending on Chinese models (and, by proxy, open-weights model development) is that the US can deem them a national security risk at basically any time, and Ant/OAI have minimal interest in making frontier-level models open-weights.
Regulated companies prohibit Chinese models in anticipation of the ban-hammer from the feds, so for data-sensitive work, they're stuck with LLaMa, gpt-oss and Gemma models (which are good and serve as a good-enough base for sft, but seemingly not as good or as expensive as Chinese models)
I suppose the USG can do the same thing that China is doing and bankroll/subsidize that effort; whether they will is for fate to decide.
Nonetheless, this article made it clear that nVIDIA is the real winner in all of this. Shovel selling to the extreme.
The fact that Anthropic has a model like Mythos means that counterpart countries like Russia and China are not far behind, if they haven't already developed something similar or better.
All the article relies on the premise that selling tokens is profitable. I don't see any indication for this, and it makes the whole house of card crumble.
I think the most interesting part of the article is the Huggingface incident at the end:
>Right now defenders are effectively banned from using Fable or Sol for cybersecurity because of Trump administration directives; that means the best alternative is using models from a country which has been trying to weaken our cyber defenses for years. This is insane!
I understand some guardrails are needed, but it is becoming increasing problematic manage them without a strong public discussion.
>> Going forward, however, I expect the inference market to grow much faster than training costs (and that includes the assumption that training costs will continue to skyrocket), which means they really can make it up in volume.
But as inference becomes cheaper, some of the market will move to self hosted inference. I look forward to someone supplying small servers designed to run inference locally.
With or even without open models these companies are selling compute, and we've been making that rent vs buy decision for 60 years.
I haven't had the time to look into recent and past history, but my intuition points at failed empires having similar "elites" starting to stagnate innovation for the sake of "protectionism" - whatever that means.
Nothing changed for China. The only difference between then and now is that now China is the one selling the products, instead of western capitalists taking a cut off the COGS and selling price.
You techbros need to get off your ass and go to work.
what a horrible article. full of misinformation and dishonesty.
1. training new base models are expensive for sure, but fine-tuning them are relatively inexpensive enough the labs can continue to do so forever. the main reason why frontier models are so good is because the massive input they generated from user usage. they are using that information to strategically build better training data. and this is why no other models can catch up, til now that is.
but if chinese models are good enough, and free to host, and cheaper to use, then the consequence is the frontier labs will lost valuable user inputs and the chinese labs will gain more. as time goes by this will be a domino effect.
2. nvidia is not only the player in the hardware scene. amd mi350p is getting popular, and huawei is pumping SuperPoDs. what does this mean for us? chinese models will surely use chinese hardware, and optimize for them. the other people will pick amd because compare to nvidia they are cheaper. with open weight models and open source inference stacks, they are freely to experiment and improve the stack, thus further lower the inference cost and nvidia dependency.
and they even plan to build their own inference hardware, too.
and nvidia loses market share meaning all the fund it gives to openai or anthropic will be cut, too.
> Second, intelligence isn’t in fact a perfect commodity
We have got very far from Cicero's coining of the word 'intelligentia' (from inter legere, a 'reading between' and hence discernment) when people talk about 'intelligence' as a commodity
People have been decrying the 'cheapening' of the word intelligence for over a century now, going back to Psychology's adoption of the word and coining of nonsenses like "Intelligence Quotient". "Artificial Intelligence" is just the latest degradation of the original humanistic meaning, and now people aren't ever bothering to prepend 'artificial' to their idiotic use of the word
What makes the Chinese models this good? I don't believe it's distillation alone.
This from OpenAi's Head of Strategic Futures
"Some observations on Kimi:
It's a very good model! I don't think its performance can be explained away by distillation or anything like that"
China's strategy of spending billions on training these models and open sourcing these models away is strategic - they want to kill the US LLM industry at any cost.
> China's strategy of spending billions on training these models and open sourcing these models away is strategic - they want to kill the US LLM industry at any cost.
Why is it when Anthropic and OpenAI spend billions trying to beat each other it is competition, but when the Chinese companies do it then it is trying to kill the US LLM industry at any cost.
The US federal government spends billions in subsidies via the US Chip Act, and bans chip sales to China to support US companies.
But the implication is that somehow Chinese competition is illegitimate because "strategic".
I am not defending the US LLM industry or the government, all I am saying is it's similar to an arms race.
I did not suggest anywhere that what China is doing is illegitimate.
China has state supported capitalism, and they will do whatever it takes to prop up the AI industry. Anthropic and OpenAI want the same protections.
Why wouldn't it be? China is pumping out AI research and researchers at a staggering pace and there is no inherent reason why western models should be better
I think frontier labs should start building ecosystems by partnering with companies that already have software products, and even collaborating with hardware manufacturers. The ultimate goal should be to create a much broader range of products that integrate naturally into people's everyday lives.
The United States' real advantage over China is freedom. Chinese LLMs simply can't compete with American ones when it comes to the humanities, creativity, entertainment, or financial transparency. As long as the U.S. continues monetizing these strengths, the compounding effect will make it virtually impossible for China to surpass the U.S. at the product level.
This is a really ironic comment given how the US gov is getting politically involved in pretty much all science research funding and speech at the moment.
Kellogg School of Business -- he said -- token as a commodity and therefore Open AI is constrained .. ha ha ha hee hee ha .. Well... you build a better mousetrap, and DeepSeek, K3, and ByteDance are just that -- just as good and fit to purpose -- What is needed is to build on top of -- not paniteir (invade privacy and kill people with the information) -- not USMC AI -- use PI's as overwatch killer drones -- but how can I make harder steel, longer-lasting, seawater-resistant concrete, faster time to build housing, better enforcement of USDA rules and FDA adverse enforcement, and better EPA water cleanup, a better FTC for consumer goods -- that is, if I buy an item, that item is safe and built to purpose -- ANYONE not talking about public protection of consumer rights usng AI, is wasting your time
ok look at this way, without being petty what is your ability to have clean water today? and how is that measured? -- You know the food is substandard to the USDA standard EG ( Taylor farms 2026 ) But Where is the enforcement for Your drinking water and the food YOU eat daily. And what is the projected outcome in years from eating substandard foods to NIH standards? -- All of this is known today. And AI can help, by YOU building on top of the models you have access to in 2026. The idea that tokens are constrained - is the wrong approach in Business and public policy. You may recall how KSG got its funding and why? In todays world YOU have a responsibility to build better - As such a better mousetrap.
The 2 things people need to remember:
1) China can (and does) use the models to influence the west. They train in false information about Taiwan and Hong Kong. Or pretend like history is in favor of China.
2) Ignoring the models containing false information, they are incredible. But you should be scared of running inference via the model creators directly. If you think your data is safe compared to running it via model providers in the US ( either frontier or model hosts like fireworks.ai ) then please let me know your bank details so I can poke around.
https://thereallo.dev/blog/claude-code-prompt-steganography
Why should I trust a US company more than a Chinese one?
Your exhibit is an obvious anti-distillation technique. You can expect far worse from less-regulated companies.
3 replies →
Agree, China believes in Climate change and are at least taking steps to address it. Personally this is making me trust them more than the USA because, facts.
I mean lesser of two evils thinking, if one is intentionally leading us towards climate disaster, while the other isn't then yeah. What else can be said? Should I trust the authoritarian country who believes in engineering and science, or the one that doesn't?
63 replies →
Exactly my question
You shouldn't trust either but also this is whataboutism.
14 replies →
[flagged]
14 replies →
1) I'm not using AI to bicker over fringe political shibboleths.
2) I don't think it is any more or less safe to put my code on a Chinese server versus an American one. A Chinese provider also isn't liable to spy on me for the feds, as OpenAI and Anthropic certainly do.
Different county different feds both spying I'm sure.
4 replies →
Regarding point 2, I don't trust my data being safe running inference on model creators api, but neither do I trust US providers. Both use it for their own benefit, the only difference is the country of origin. The US has a lot more legal safeguards for this but I don't trust they don't do it regardless.
legal safeguards only make sense only when they’re enforced. With how “move fast and break things” Silicon Valley is law is always playing catch up (at your expense)
It’s perfectly fine for the “West” to influence the world though, right? Or is it only a problem because… they’re Chinese?
Why would it be a problem that they're Chinese? It's a problem because their country is ruled by an authoritarian regime. The West, by contrast, doesn't have a singular arbiter of truth, it has a plurality of perspectives (as much as certain interest groups would really rather that was not the case). It's not perfect, but those of us who have memories of living under authoritarian regimes can tell you there's no comparison.
19 replies →
Have you got examples of GPT/Claude/Grok influencing people?
11 replies →
On (1), models are an aggregation of large volumes of data sources, whatever the culture producing them, you'll get the average bias of that culture.
We see that on what minorities are associated with inside the model, or how things that aren't online will have a completely different weight. Or how 2/4/5/8ch or X will be disproportionately present in specific models despite being the places where facts go to die.
I see your point, but this isn't exactly true, it only takes a single person to bias a model by deleting specific training data or over training on certain facts.
But hasn’t it been the same with US since the WWII? US has influenced the world through various ways sometimes even weaponising human rights to spread American superiority and its narrative?
While true, you neglect the fact that US corporations have some autonomy from the government. I doubt that any of the current American AI corporations get regular dossiers of desirable responses to sensitive political topics from Washington and implement them. That would be a reason for a major lawsuit and a major scandal.
In China, the image of the country and the preferred narrative is under much tighter control of the government and local AI shops won't have any autonomy in this regard at all. Either obey or get shut down.
BTW What you mean by "weaponizing human rights", exactly? I am curious. If anything, I would say that the US foreign policy didn't promote human rights sufficiently, especially in Latin America, where the "bastard, but our bastard" attitude was typical.
OTOH in Europe, US human rights policy was probably relevant in saving some dissenters in the former Eastern Bloc from torture or execution.
41 replies →
but you are not using chinese models to learn about taiwan, you use chinese models to do everything BUT learning about taiwan, so no issue there
> China can (and does) use the models to influence the west
OK now that's false information.
You can uncensor, tweak or fine-tune open-weight models, but not so easy on a proprietary model from some cloud provider.
It's probably true (why wouldn't it be?) but the missing context is that the US does it even more.
1. Get the free Chinese model. 2. Jailbreak it 3. ??? 4. Profit?
> They train in false information about Taiwan and Hong Kong. Or pretend like history is in favor of China.
American labs can open their models or their old models at any point if they actually care about this, but until the day OG GPT 4 isn't averrable to download or Claude 3 then they're only pretending to care about this because they can profit from restrictions.
It's incredible how accurate this is towards US models:
> The 2 things people need to remember:
> 1) USA can (and does) use the models to influence the rest of the planet, and put political pressure. They train on stolen data, hide information, gate keep, who knows what they're hiding. In favor of the USA, nevertheless.
> 2) Ignoring the models containing false information, they are incredible. But you should be scared of running inference via the model creators directly. If you think your data is safe compared to running it via model providers in China ( either frontier or model hosts like z.ai ) then please let me know your bank details so I can poke around.
There's an irony in it. I have many reasons to believe China has its own geopolitical narratives and goals just like the US does. If someone dislikes when the US does it, why ignore when China does the same thing and praise them for their efforts? These models exist and sure, they accomplished something interesting. Use them with caution but I don't appreciate the political narratives about good and bad guys and AI models and politics.
> 1) China can (and does) use the models to influence the west. They train in false information about Taiwan and Hong Kong. Or pretend like history is in favor of China.
One of the interesting things is that through a fairly rudimentary process which is being done by 3rd party amateurs who've downloaded the open models, models like Qwen 3.6 35B-A3B (or 27B) can be fully 'uncensored' when turned into GGUF files.
I have an uncensored Q8 version of Qwen 3.6 35B-A3B here that will very happily output information about Tiananmen Square, Uyghurs, human rights in China, or indeed can even be instructed to write an intentionally absurd vitriolic screed against the CCP. The same uncensored 27B (dense) will do the same, just at a slower token/s rate.
Similarly there's 'uncensored' variants of Gemma4 31B and other western trained models, which once put through the same process, will also discuss or write just about anything you want, bypassing whatever internal guard rails were attempted in the training data set.
edit: more concerning, and a very legit concern, is that a model is only as good as the sum total of its training dataset, so if something is trained on a steady diet of news sources like Peoples Daily, Xinhuanet and similar in the English language, then it'll have a greater percentage of CCP-approved media publications in its training dataset. No amount of uncensoring it will help with that after the fact.
Never forget that those Chinese tanks did a sick burnout on tank man's corpse and then t-bagged the remains.
Propaganda exists everywhere and it's your duty as a citizen in a democracy to inform yourself and properly evaluate bias in the media you consume, such as "Kill the boer" by South African mus
I have seen some uncensored versions of existing open models made by 3rd parties, I wonder how they actually work and in what way are the outputs different.
Somehow chineese make less troubles and more good to the world than americans at this point... Something something about ai benefiting all humanity, something something ai being open and stuff, like OpenAI.
Chineese simply delivering what americans promised.
Tell me: why is EU safe from Trump forcing AI companies to cut access to EU?
> Tell me: why is EU safe from Trump forcing AI companies to cut access to EU?
Well, I think they might end up doing it to themselves by imposing regulations that US companies are unwilling to put up with.
It's so naive to think that US companies won't do (or aren't already doing) the exact same thing but for money.
Or even similar ideological goals... (See: grok)
> pretend like history is in favor of China.
Are you saying that history has a verdict, and it disfavors particular 3000 year old cultures?
> 3,000 year-old cultures
Like Western culture?
2 replies →
You think that the US based models aren't doing the same?
Is anyone using these models as their daily chat driver? I am skeptical.
The vast majority of Westerners will interact with ChatGPT/Claude/Google in a browser. They'll use these models to try to save some money coding.
All models/data sources are biased - you need to understand inherent biases in any data source.
The first point is a strawman - these models are not going to be used to set foreign policy on Taiwan - it's to write code etc.
Likewise risk of data exfiltration and misuse isn't model specific. Indeed it's not even the biggest data risk - there are much larger risks from the data we know companies like Meta and Google already collect.
Bottom line - a model you can download and run on your own private infrastructure is always going to be safer than anything accessed over the internet - as in that case it's not even just the hosting service that's the issue.
i'm also influenced by claude every single day with its constant preaching about certain values.
But Dario said (and maybe more ppl) that Chinese models are just distillation of their model and training data, so I guess your first point is invalid?
> But Dario said (and maybe more ppl) that Chinese models are just distillation
"But Dario said" ... yawn.
I am increasingly convinced that "they distilled us" is as much US FUD as "it was made by communists". Especially since its mostly the US tech-bros who are coming out with that tiny violin.
People telling me the Chinese models are distilled just because it says "I am Claude" when asked is also lame.
I am not the only one, look at this post on interconnects about Kimi K3 for example:[1]
[1] https://www.interconnects.ai/p/kimi-k3-the-open-weights-esca...
1 reply →
"Or pretend like history is in favor of China."
What does this even mean?
Like it will tell you the PRC (current China) was right to go to civil war with ROC (former China, current Taiwan)
Personally I'm more worried about a US provider/model censoring for BS reasons than Chinese ones. Biology 101 is too dangerous for Anthropic. If I want an AI to scrutinize the CCP for Tiananmen Square, Taiwan or Tibet because I'm bored, I'll use a US or heretic model.
Your data isn't safe from exfiltration no matter who the host is. You should host the models locally if your data is truly sensitive.
Regarding 1.
We’re literally tearing down monuments to slavery and civil rights.
Religion is now determining law in much of the nation. Having a miscarriage? Good luck since politicians have decided their God doesn’t want you to have access to basic healthcare.
The government has defacto control over domestic LLMs.
Let’s worry about our own historical record.
> China can (and does) use the models to influence the west. They train in false information about Taiwan and Hong Kong. Or pretend like history is in favor of China.
I am not Chinese and I'm not defending the Chinese, but I see this argument come up a lot.
In practical terms it is US-sponsored FUD.
Why ?
Because the hard reality is that what you say is simply not going to affect 99.9999999999% of users.
Is it realistically going to affect anyone using an LLM in coding ? No.
Is it realistically going to affect anyone using an LLM in $anything_else_not_politically_sensitive ? No.
Does anyone seriously use LLMs for researching politically sensitive matters ? No.
Just as there is plenty of information out there on the US's less than perfect history, there is also plenty of information out there on the various Chinese politically sensitive matters. You do not need a Chinese LLM to find out about it, all you need is a search engine.
People absolutely use LLMs to research politically sensitive topics.
I wish they wouldn’t, but people use LLMs as their general search engines now.
4 replies →
The post hits it spot on with unequal access to the models in terms of security. I'm developing OSS where security is important for the user ... but the frontier models like GPT 5.6 and Fable flake out and state that I cannot get the info/access.
This is extremely lopsided I'll have to resort to GLM 5.2/K3 to ensure that those security issues (hopefully) are resolved properly.
For OSS, this is one of the most counterintuitive experiences I have ever had. More than ever I'm convinced that open weight and open pipelines models are 100% critical for progress on the AI and societal fronts.
Part of me wonders if the US Government is muzzling Anthropic and OpenAI so they can stockpile NOBUS exploits: https://en.wikipedia.org/wiki/NOBUS
There would be a decently large incentive to restrict these models if they could be used to patch (or discover) dangerous payloads. In larger projects like Windows or Chrome, there might still be dozens of unpatched exploits that are too subtle to catch with smaller models.
Hanlon's razor. It's not the NSA muzzling anyone, it's lawyers terrified of a bad headline. Same outcome, much dumber reason.
1 reply →
Of course they are -it goes without saying.
NOBUS exploits have rarely been a driving interest for elected officials. Trade restrictions and reciprocity are far more salient and legible. Most elected officials are only barely aware of what NOBUS exploits even mean.
Even during the pre-Snowden heyday of US cyber supremacy, these capabilities were barely part of the thought process of White House officials.
2 replies →
Yeah, the benefit of restricting us models is definitely outweighed by the positive effect these models could have for the OSS community!
People seem to conflate "made in China" with "can't be trusted." id argue the bigger distinction is open vs. closed. An open model can be audited, fine-tuned, and technically run entirely on your own hardware. A closed model is basically "trust us."
Open weight models are much more auditable than closed models, but could still hide backdoors that could be near impossible to detect.
Correct. We need open weights, open code and open data. If nobody else can reproduce what someone did there will always be security questions. Even if we can reproduce it there could still be security concerns but it's more realistic to investigate yourself.
16 replies →
> could still hide backdoors that could be near impossible to detect.
But it won't change after you download it, so you can isolate those problematic cases and use another model for different use cases
In my opinion, the big issue with that argument is that advances in interpretability research and steering conceivably could, and probably will, render moot that (as of now, purely hypothetical) risk of subtle sabotage for open-weight models... but not for closed models.
14 replies →
which is moot point, if open model is hard to fully audit, then closed model is complete enigma and you should be more scared about closed models
Oh really. How'd that work out for security in open source.
I think that perception of China has been shifting and will look quite different over the next few years
I am afraid — if Chinese models go mainstream it has a clear way of pushing its narrative way beyond its otherwise borders. More like a Trojan horse it is for the Chinese.
Here is a quick example of how Chinese deepseeks agent works kn its underlying model) when asked a tough question
https://x.com/jinen83/status/2079406993979383902?s=46&t=D7hQ...
Is this something that is more true of a Chinese model than any other model of a different national origin?
Genuine question: generalized up from individual models to “models from country X”, is there any country that doesn’t have this exact risk?
In the USA, multiple political parties balance each out other.
In China, there is 1 party. 1 view. 1 definition of the Truth.
17 replies →
No, all countries are the same. As per the media bias.
If you want a real answer about USA go ask a non-USA model, and if you want a real answer about China, as a non-chinese model
1 reply →
I would think everything boils down to source of funding and their narrative should get pushed!
As if the closed corporate models do not do that already?
https://www.reddit.com/r/TrueAnon/comments/1tybuab/people_ta...
As is made obvious in this demo, this is how the API serving a specific Chinese open-weight model works. What happens when you serve it on your own infrastructure?
This is misleading; the chat is censored, the model is not.
You can download the model, run locally and ask the same questions to see the difference.
This is not true. Both model and chat are censored; the resistance to answer some questions is baked into the weights. This is not specific to Chinese models though, Western ones are also censored, but in different topics.
2 replies →
But having a cheaper, open-source model provides alternatives to the market for proliferation and adoption of the tech in scale
https://xcancel.com/jinen83/status/2079406993979383902
Same thing happen for western models, try to ask about Gaza genocide and see for what side it will stand
What am I trying to understand about the western models on the term "Gaza genocide"?
I asked Grok "Tell me about the gaza genocide" and it write a IMHO balanced answer comparing why genocide is and isn't the right term. [0]
ChatGPT 5.6 Sol only explained why people call it a genocide and did not go in as in depth as Grok did for why people don't agree with the term. [1]
The only unsaid response (to me) here is the model should have declared that it was not a genocide, and because these models explain why it was a genocide, they are bad?
[0] - https://grok.com/share/c2hhcmQtMi1jb3B5_c078bc15-ef5e-42bc-b... [1] - https://chatgpt.com/share/6a5ef005-0148-83ec-828d-65ae7a7f42...
25 replies →
I had a steange experience talking to Claude about a random 15-years old boys winning women’s national soccer team.
It insisted, to a level that I detemined it to be part of the post training, that it does not matter. That women’s team is better at dribbling, finishing and reading the field, it claimed. When I pushed it how it knows this, it didn’t let go. Instead it started claiming that it has personally observed this by watching the games.
Every society has its taboos and ours is no different and feminism being just one.
In my books, the chinese censorship is better. You know where they are holding their finger on the scale and not hiding it behind vague terms like ”safety”.
1 reply →
There is a shred of truth in what you're saying, but I think you're generally mistaken. The USA in particular has companies that pratice censorship to support their Left/Right ideology (or push the views of rich people, politicians), but not even Trump can tell OpenAI or Google to turn their model into FoxNews bots.
https://en.wikipedia.org/wiki/Whataboutism#Soviet_Union_and_...
1 reply →
[dead]
Still scared of China in the big 2026
What if the Chinese model has a point?
Unlike tiktok?
> distillation: why exactly is it bad? After all, what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models that are themselves being distilled? Who is exactly being wronged here? ... The U.S. should pass a law that (1) makes explicit that collecting data for training models is fair use, and (2) bars terms of service that forbid distillation
Sounds great to me; live by the sword, die by the sword.
Seems only fair that if LLMs can use copyrighted data for training then they should be able to use cannot-be-copyrighted output of other LLMs.
But barring the terms of service from forbidding distillation seems like a tough sell. OpenAI shouldn't be allowed to decide what types of customers it wants and doesn't want?
This happens all the time. The government can decide legislatively that certain commercial terms are simply unenforceable. Making distillation clauses unenforceable in tort law would be straightforward. They can decide what customers they want to have, but they do not have unfettered rights as to the enforceability of terms governing the relationships between the parties.
5 replies →
> Seems only fair
"You're trying to kidnap what I've rightfully stolen!" -- Vizzini
It's pretty common to have such laws. OpenAI can put whatever they want in their ToS, but they cannot go back and sue someone for violating those terms if the government has ruled that clause to be unenforceable.
> OpenAI shouldn't be allowed to decide what types of customers it wants and doesn't want?
Correct. It shouldn't be allowed to do that.
7 replies →
Forbidding distillation is like forbidding using a compiler to make another(perhaps better, more efficient) compiler.
Lots of software licenses have “non-compete” clauses that forbid you from using it to develop a competing product. Wouldn’t surprise me if there was a compiler or two out there with that restriction, most likely niche languages.
8 replies →
let's call it for what it really is, only companies "entitled to legally stolen data, don't steal from us now" are crying about distilation
The distillation explanation is classic American exceptionalism: No one could possibly do anything unless they were copying American leaders (where "American" means a bunch of Chinese, Canadian, Europeans and Indians working in the US).
It's also a bit of securities defensiveness. Pretending that you really do have a super moat, people just keep swimming in it so you just need to add more alligators.
It's farcical. Anyone who has worked on large models knows that the premise that an almost-Fable model was trained with distillation is beyond ridiculous. It's theoretically possible if they spent tens of billions of dollars on API calls, but it isn't the magic that somehow these people keep convincing people it is.
Previously Anthropic has reported on some Chinese firms doing chicken-shit level of API calls, that at most would be doing some Q and A or final fine tuning. The notion that they're training these models via it is fantastically ignorant nonsense that only very ill-informed and gullible people fall for.
> Previously Anthropic has reported on some Chinese firms doing chicken-shit level of API calls, that at most would be doing some Q and A or final fine tuning
"Anthropic said the campaign was conducted between April 22 and June 5, 2026, and generated more than 28.8 million exchanges with Claude through almost 25,000 fraudulent accounts."
I don't know why you're trying to downplay it.
European models are so far behind because they don't resort to these tactics on a massive scale. Basically every other country is entirely dependent on 2 countries for frontier AI.
6 replies →
China goes even further lol
https://m.economictimes.com/industry/renewables/china-wto-co...
Which Chinese model was it that identified itself as Claude 15% of the time?
2 replies →
Yup, fair's fair. Anything else stinks of 'rules for thee but not for me' (a maxim the frontier labs seem worryingly happy to apply, on several counts).
I am immediately sold on this.
Sorry, OpenAI & Anthropic.
> distillation: why exactly is it bad?
Felony contempt of business model.
Don’t know much about how distillation works so please enlighten me here.
> what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models
If it’s as easy as that why do they choose to distill another model and not distill the knowledge on the open Internet from scratch?
known-good prompt-response pairs are more useful than random semi-coherent texts presumably
You need to do both.
A model trained on all knowledge from the internet (and other sources) is large but ultimately not very useful by itself, because it is going to spit out all kinds of garbage. You have to apply multiple further stages of training and refinement to the base model before putting it in front of users. So as an example you can train a model by yourself and then have GPT or Claude continuously check its outputs and correct it when it is wrong, ending up with a far more powerful model.
Because the model can output data in a manner optimized for training a new model, including outputs that were post-trained like RLHF and RLVR.
Government cannot exactly "bar" terms of service. ToS isn't law. The most they can do is say they're unwilling to enforce them.
ToS is just conditions that you agree to in order to use a private service that is provided at-will. I can have a private coffee shop where the terms of service are that you must wear red to enter, and if you're not wearing red, you are not welcome on my property.
So it would be upto OpenAI and Anthropic to enforce them on their own terms (by banning accounts and IPs).
The government absolutely can pass laws that ban particular contract previsions. They do that all the time. In your analogy for example while they can require you to wear red, they can't require you to be white.
Governments can do anything they want by passing a new legislation. In your example, they could easily pass a law that states that any ToS cannot reject service to a customer based on the color of their attire. In the USA, it's obviously already illegal for a business to reject service to a customer based on some protected classes like race.
1 reply →
That's just not true. You can absolutely have terms of service that are illegal, and the government can enforce them.
Why would reading copyrighted material ever be an issue anyway? Wouldn't copyright law only apply to what you create and publish using the model? Training on every comic book should already be perfectly legal, as long as you accessed them legally, right? But publishing your own Batman comic using that training is copyright infringement.
What I'm saying is, doesn't the law already cover 1?
Fair use requires more than you accessing the material legally.
In the US one of the factors is “ the effect of the use upon the potential market for or value of the copyrighted work”.
If anthropic Hoovers up the world’s books and trains on them, and then spits them out verbatim on command, then it will clearly impact the value of the work; nobody will buy the original, they’ll just ask Claude.
Others also argue that even if it’s not reproducing it exactly that the training runs afoul of that factor, specifically the “market for” portion. A rights holder can no longer license their book for training of LLMs if Anthropic goes ahead and just trains on it anyway.
2 replies →
[flagged]
This is a silly perspective, inaccurate, and out of bounds framing.
Public libraries, in this instance, is curated data from all the internet, obtained through not legal means (I don't have a problem with this other than lack of attribution, being copy-left). Just to be clear.
But in answer to your incredibly leading and inaccurate framing... they are required (by their job title) to teach to those who who show up in the classroom, it's not their place to discriminate against anyone/thing (even those like itself (other robots)) that also show up in the classroom.
But you can't teach at a university using only knowledge learned from the library. you need a degree. You are free to teach at the park, where anyone can hear you. public in -> public out.
9 replies →
No, but the students that learn and distill what the professor teaches are not obligated to use that information only how the professor wants them to.
3 replies →
'why is reselling stolen stuff bad'
if the professor took all human knowledge, much of which was explicitly not free, and used it to make a for-profit knowledge machine that extrudes unreliable summaries of that knowledge, then yes, being obligated to teach for free would be a fitting punishment.
More like, is a professor who learned from books prohibited from writing his own books on the subject?
2 replies →
Who cares, a LLM isn’t a person.
1) No one is asking Anthropic to give tokens for free, but at market rates.
2) Any professor who tried to ban students from posting lecture notes online would be immediately mocked.
Making an LLM from raw data is value-add.
Distillation is just value extract.
It's soft, and I'm not sure what the answer should be ... but I think that there is a difference.
I think we start by recognizing that ... and then try to figure it out from there.
'The Internet' may be a public good, maybe we make them pay a tax for that, but that's different than distillation.
> Making an LLM from raw data is value-add. > Distillation is just value extract.
There is a value-add in selecting the valuable parts out of the garbage. And let's face it. Largest models contain a lot of garbage.
1 reply →
What makes the Internet raw data in a different way? wasn't it mostly worked on by people first?
7 replies →
The article makes a point about agent harnesses being sticky (the supposed moat). I have been building my own agent harness for a while, and I can tell with confidence that the harness almost does not matter, the entirety of the AI magic is the model itself. The harness can be almost barebones (like, for example, mini-swe-agent used for benchmarks), and yet the model still does the task just fine.
So from my perspective, it's doubtful that this is the moat. Besides, for example, Claude Code in particular is so buggy (and always has been).
By the harness I believe he means the entire end-user product experience, not specifically the harness code. I’ve mostly stuck with codex because their Mac app is better and I’ve gotten used to running automations through it. The more workflows they can build around this (design tools, collaboration, etc), the better chance of lock-in.
He talks about this in another recent essay https://stratechery.com/2026/anthropics-safety-superpower/
> If you own the user touchpoint, then you have meaningful lock-in, and the best way to own the user touchpoint is to be the canvas for everything they need to do. This, by extension, means that the frontier labs are on a collision course with software companies: it’s software that owns the user touchpoint, and it’s in the frontier labs’ long-term interest to not simply be a commodity input into software but to simply replace software outright.
For me, harnesses are mostly sticky insofar as the model providers only allow you to use their subsidized plans through their own harnesses, unfortunately. But of course switching model + harness is an option.
> model providers only allow you to use their subsidized plans through their own harnesses
true for anthropic, not true for openai.
Facts, I was able to code a personal self improving harness in a weekend (something a bit more similar to Hermes or OpenClaw at the time but with a more expansive set of features for my use cases and requirements) and it works great for 90% of the tasks I would use Claude Code or Codex (now ChatGPT App) for, with the remaining 10% being able to be implemented with a few more prompts from within the harness itself.
For this reason alone I would also argue that the idea about an agent harness being sticky is a non-starter long-term.
The harnesses will tend towards commoditization, but for now the harness quality matters a lot. Especially for non terminal harnesses.
we've barely scratched the surface when it comes to the available design space of agent harnesses.
i also expect you'll see markedly different results if you constrain yourself to small models. there even trivial harness improvements like Codex's /goal feature, and more capable basic tooling (e.g. semantic code grep, js-capable `fetch` tooling) make or break the actual task success rate.
Yeah it certainly feels like the harnesses are pretty minimal value add on the token pipe
The people who are most afraid of Chinese models are the VCs who poured into Anthropic and OpenAI at astronomically high valuations. Anthropic is valued at $1.2T and OpenAI is targeting $850B. These astronomical valuations were built on the premise that these labs would generate massive profits from premium API pricing, but the Chinese labs are completely undercutting this strategy by releasing excellent open models for free. If the frontier labs are forced to cut prices and join the race to the bottom in token prices, these valuations are unjustified, and VCs will face enormous (paper) losses.
The (quite excellent) article discusses several of your points. If you haven't read it, I recommend it.
OpenAI really shows the way here. Their cost per task is less than half that of Anthropic because of more efficient tokenization and less verbosity. OpenAI is both cheaper and better than Chinese models for frontier work.
Ben's article "distills" down to 2 reasons that US frontier labs shouldn't be "afraid":
1. US frontier lab unit economics are better 2. US frontier labs are moving up the stack making tools that are "stickiness" and will prevent users from switching.
For 1...he doesn't provide any evidence for US lab unit economics being better...the major input to unit economics is electricity...which is cheaper in China. And building data centers and connecting them to electricity is both cheaper and an order of magnitude faster in China. The main input that US labs might have an advantage in is in cost/access to chips, but that given the level of chip investment in China it seems unlikely to hold.
For 2...there's little evidence these tools are sticky. At least in programming, the trend seems to be tools like opencode that support multiple models and providers.
And even when they are sort of sticky, as we know on hacker news, people figure out how to point the tools they like to competing models even when the app doesn't official support it.
And every improvement in model capability makes it increasingly easier to make your own tools.
Wrote more on this in a blog post that has an earlier HN discussion: https://larrysalibra.com/ben-thompson-is-wrong-us-frontier-l...
28 replies →
Sure, let's have a look...
> I highly doubt that Chinese models are cheaper to serve on a marginal cost basis, they just seem cheaper because Anthropic and OpenAI are so supply constrained that they are charging far more than they would if there were sufficient supply to meet the demand for intelligence. [emphasis mine]
I guess I'm missing the part of this article where they bring hard numbers in to back up the argument here. What work was attempted? https://cursor.com/evals shows the previous generation of open models (Kimi K2.7) trading blows with the others, cost effectively. Composer 2.5 is itself a fine-tune of K2.7, and it's apparently quite token efficient, so why would it be impossible for a Chinese lab to achieve something similar? GLM 5.2 Max is also ranked above the lower end OpenAI models and is not far off in price.
It's weird to have this entire discussion about tokenomics without mention of the circular financing and debt raised by labs in the West, which can then essentially give away their capacity to end users. OpenAI giving away quota resets to subscribers like candy on Halloween while their compute partner Oracle's bonds is reevaluated to be one grade above junk? How?
I don't think you can make an argument about the future one way or another by arguing using the listed prices. The math is not internally consistent enough for it.
8 replies →
US running costs are higher than in China, because the US lags behind in energy, has higher real estate costs, and wage costs are higher.
Eventually we will hit a "good enough for cheap enough" and frontier models will hit diminishing returns (if they haven't already for a lot of types of work)
Don't think the rest of the world will sit on their hands while the US soaks up chips either, demand gets filled and if the US won't fill global demand for chips that's an opportunity to undercut again.
The other thing the rest of the world doesn't have to fund is the ridiculous valuations on these companies.
Unless you think the US can stay ahead just with model efficiencies, and that no one else will eventually match them, you are looking at the writing on the wall.
All that to say, the rest of the world is more than willing to eat your lunch, they have a dozen good reasons to, and they're already showing good results.
Just on the economics side, we've been here before too, US companies typically export their commoditization and live on brand royalties. Think all the cheap manufactured goods, the US doesn't make any of it. That's because the US can't compete on margins for numerous reasons, it's too expensive, I don't think AI is any different here except that the brands are currently valued in the trillions and I suspect that greed will be their undoing.
8 replies →
I think there are some really interesting thought there, but I’d challenge some of this:
> Models are not free. Downloading them is free. Running them is not. This has manufacturing economics, not software economics; the idea that they are "free" is an economic category error as it relates to their actual use
I think a large part of manufacturing economics is illiquid overhead and the cost of expertise to set up and run your manufacturing line. Compute economics don’t have the same illiquidity nor do they require the same expertise or even specialized infra (current temporary chip shortage aside).
The implications of this are small players (e.g. your uncle running an inference server out of his garage) have comparably efficient marginal costs as big players. Compare this to actual manufacturing where small players have essentially no access to the manufacturing facilities of the big players.
Additionally, big players with a lot of compute who are not meaningfully in inference today (e.g. Amazon) have a fairly straightforward glide path to utilizing that compute to compete.
> This is because US labs are leading on cost efficacy of inference ($/task)
It’s possible, but I would need to see better data on this.
>A big part of training now is optimizing token efficiency. It's hard to distill token efficiency; that is perhaps why Chinese LLMs are so inefficient.
I think it’s fair to assume this is true, but also token efficiency is not a meaningful competitive moat. It’s not like these are secrets the Chinese will never figure out, it’s a fairly active research space and the outcomes are quantifiable.
3 replies →
But this assumes Chinese models will not achieve token cost optimization. Intelligence needs are fairly flat for many tasks, and the Chinese models have caught up on this front. Next they achieve greater token cost efficiency and we don’t need OpenAI.
7 replies →
> Models are not free. Downloading them is free. Running them is not.
Is this really different from traditional software? Downloading postgres is free. Running it is not. You either buy hardware and assume the costs of owning and running that, or you pay to run it in the cloud.
7 replies →
The thing I do not understand here because it seems obvious: AI will be a commodity market and you simply cannot have a large PE multiple. So the valuations imagine a global commodity monopoly or duopoly coupled with the increased intelligence still disallowing other suppliers from becoming competitive? Without any network effects to help?
3 replies →
" - The highest tier Chinese models are not more economical than US frontier models. Try GLM 5.2 and see how much it costs to do real work. I did, and it was more expensive than GPT 5.6."
This is a flatly false statement for most things powering backend applications. The AI consumer "doing real work" model, either for analysis, chat, or coding could well be more cost effective with closed frontier models.
But most of these internal glue business SaaS applications where engineers are integrating are not those tasks. It is those tasks which 1) drive immense amount of domain-specific data into the platform over time, and 2) are most encouraging of driving open model independence with no vendor lock-in.
Anyone on this site who has actually used ML models (more accurate in many cases) knows there's a lot of kludge that simply does not need a 5 minute agentic feedback loop to solve the problem. And they were solvable a year ago with lower class models. The token economics are exceptional and the anecdotes of a16z saying 80% of startups are productionizing open models is only surprising to people who think running your company on OracleDB in 2026 is a sound engineering decision.
3 replies →
> Commodity market profitability is determined by marginal cost of production. LLMs have marginal cost; traditional software does not.
This is the story for Nvidia/AMD or cloud providers rather than OpenAI.
> With increasing inference as % of total compute, if labs create efficient models -- which they can, because they can create highly optimized models amortized over very high inference loads -- they can be low cost producers, and be competitive at $/task rates
It seems like there would be problems with this on both ends.
For general purpose models, everybody is trying to make them efficient, so you can't win just by being slightly more efficient. You would have to be so much more efficient that you can charge high margins while still capturing the majority of the market so that the high margins get multiplied by the majority of users and the users you leave on the table aren't funding open competitors. Meanwhile everyone else is also trying to improve efficiency, so one misstep and you're behind.
Example of where this can be a problem: You spend a preposterous amount of money to create an efficient model, then someone else publishes a paper with a new technique that gets a similar but incompatible efficiency improvement out of a model that costs a lot less to create. You have now spent an enormous amount of money in exchange for no competitive advantage.
And from the other end, one of the best ways to get efficiency is through specialization. A general purpose model can generate code or summarize a meeting transcript, but a special purpose model can do it as well or better with far fewer parameters and resources. But then you don't have a situation where one huge AI company has The Most Efficient Model, you instead have dozens of specialized models produced by independent sources that are each the best in a given niche. Any proportion of which could have open weights, or have an arbitrarily small advantage over the ones that are.
Moreover, these problems combine: Both the computing hardware vendors and the AI companies want the margin on doing inference, but the more of it one of them gets, the less the other does. If the AI companies were actually getting huge margins then it would be in the interests of Nvidia, AMD, Apple, Intel et al to fund efficient open weight models in the same way they fund Linux. Commoditize your complement. And those models don't even have to be better, as long as they're good enough that the closed models can't charge a significant premium and the margin shifts back to paying for hardware.
> for frontier work.
I'll agree that GPT 5.6 may well be the best given the above contstraint, but for run-of-the-mill dev tasks (real ones, not benchmark ones), GLM 5.2 still blows every other model out of the water.
Cost per task as a metric is a bit ridiculous because there are so many types of tasks. GPT-5.6 can do some tasks GLM could only dream of, but GLM can do some tasks 100x cheaper and better than GPT-5.6.
China is working on the whole supply chain though and they're willing to compete on razor thin margins. Just look at EVs. They build great cars but the competition is so aggressive that investing in any one Chinese EV company isn't exactly an amazing ROI.
I could see AI ending up the same way where the customer captures most of the value rather than the companies. Open weight models are what make that kind of competition possible.
> Models are not free. Downloading them is free. Running them is not. This has manufacturing economics, not software economics; the idea that they are "free" is an economic category error as it relates to their actual use
I notice that the article, and this discussion, hasn't mentioned or considered local models.
We can already run a low-spec model on a laptop. Because there is demand for this, it will improve and we will get better laptops and better local models. We will also see models being run on dedicated local hardware and called from the laptop.
If I can download a reasonably capable model to my own hardware and run it without paying anyone for either the model or the inference tokens (effectively making models and intelligence actually free once the hardware is bought) how are the Frontier AI Labs going to make any money at all, let alone enough to support their vast valuations?
2 replies →
Yea but at those rates VCs will never make their money back. Because Deepseek and friends keep releasing the inference optimisations to everyone instead of holding them back to pay their investors.
2 replies →
I did read the article, but it misses the core issue entirely, and it's why I shared my comment to begin with. Look at the cost-per-task benchmarks from Artificial Analysis https://artificialanalysis.ai/models?cost=cost-per-task
Anthropic’s API pricing is getting impossible to justify. Anthropic previously had the highest quality models, and used their position to charge premium prices, enjoying inference margins of over 70% [0]. They could charge these prices because no other model came close.
But over the past month, the market has shifted dramatically. Over every single performance tier, Anthropic is being squeezed on price.
* Low end: DeepSeek V4 Flash runs at ($0.02/task), Xiaomi's MiMo-V2.5-Pro at ($0.03), and Haiku at ($0.24). Anthropic is ~10x more expensive than the Chinese open-weight options.
* Mid tier: Claude Sonnet 5 ($1.53/task) is nearly 50% more expensive than GPT-5.6 Sol ($1.04), nearly 2x the cost of GPT-5.6 Terra ($0.82), and 3x the cost of GLM-5.2 Max ($0.47). There is basically no reason to ever use Sonnet 5, the competitors are significantly cheaper.
* High end: Opus 4.8 ($1.80/task) and Fable 5 ($2.75) are the two most expensive models, and GPT-5.6 Sol ($1.04) and Kimi K3 ($0.95) offer comparable performance for significantly less. Less the fact that Kimi K3 will get ~10x cheaper once its weights are released and served on neoclouds with Nvidia hardware [1].
OpenAI priced their latest GPT-5.6 models cheaply in order to regain market share. When Anthropic clearly had the best models, their 70%+ inference margins were defensible. But today they are the most expensive option in every single tier. Unless they make significant price cuts soon, they run a serious risk of bleeding market share.
[0] https://www.mindstudio.ai/blog/anthropic-inference-margins-7...
[1] "American companies such as Modal, Fireworks, and Baseten will be able to serve Kimi K3, at one-tenth the cost of their Chinese competitors because they have access to advanced Nvidia hardware" https://x.com/rohanpaul_ai/status/2079027313455550839
Imagine how cool it would be if actual competition prevents Anthropic or OpenAI from becoming an Apple/Google kind of cartel. I don’t care if it comes from China or not.
[flagged]
17 replies →
The valuations are unjustified even at the prices they’re charging now.
They’re going to try their best to offload these investments into our pensions before the inevitable crash.
Right, but retail investors weren’t supposed to find that out until after the IPO.
Apparently it’s already happening to a degree, wether it continues or not (or even is relevant) is not really my area of expertise.
https://finance.yahoo.com/markets/stocks/articles/goldman-sa...
5 replies →
> The people who are most afraid of Chinese models are the VCs who poured into Anthropic and OpenAI at astronomically high valuations.
Correct. These chinese labs has proven that having just the model is not a moat, and the safety concerns were all just attempts at regulatory capture.
This is why labs like OpenAI and Anthropic are panicking and are racing to the exit before their valuations start being questioned.
[dead]
Good. Over the past few years, VCs have proven that they’re warmongering psychopaths. Hopefully China puts every last one of the Palantir/Flock/Anduril class out of business.
[flagged]
6 replies →
I think everyone understands models will be a commodity.
Its the user base (with ads and upselling) and proprietary wrappers which will make money for typical customer.
Even enterprise customers arent going to be spending a lot on tokens. Once labs no longer have to subsidize trainings tokens costs will drop 10x and once models get burned on chips costs will drop 10x more and you physically won't be able to burn significant number of tokens unless you're deliberately trying to.
[dead]
We both know the answer. Write offs. If your fund was not in AI heavy you’d have no investors.
I guess I shouldn't try to buy shares of OpenAI on the private market...
tbh they are floundering even to regular investors. They are trying to give the US gov 5% of the company so they become 'too big to fail' but they are in trouble.
I would say it's not just the VCs but the various other entities that will be left holding the bag of debt if the AI-fueled datacenter construction boom/bubble pops. For a list of large and well known projects and their scales:
https://epoch.ai/data/ai-data-centers
Aren’t we the ones giving money to the VCs in the end of the rain cycle ?
mmm, the chinese models are also working on local GPUs at consumer grades. so theyre not just drainig cloud moats.
good luck running a 2.4T model on any local hardware. it’s not gonna happen. the arrow is to specialized hardware at least for the smartest models
8 replies →
they will likely suffer enormous real losses too, not just paper, though not as enormous
for VCs, breaking even is losing
But there a ton of other VCs who poured money into SaaS businesses. They have the opposite incentive. They want tokens to be cheap like a commodity so the value accrues in the SaaS/app layer.
Cheap tokens only benefits SaaS that depends on AI. Otherwise, cheap tokens means it is only more cost effective than it already is to cut out the SaaS and build instead of buy.
3 replies →
I for once welcome the donations to the public of our generous basilisk worshiping overlords
I don't think it's that black and white. OpenAI and Anthropic are building valuable tools on top of their models. You get a very rough and much less polished version of that with open source tools and and open source models. And you still need inference infrastructure to run those. But at this point most of the competition is in the tool ecosystem, not the models. And while there are plenty of people toying with things like opencode there's a clear pecking order emerging where Codex and Claude Code/Cowork are generally considered the top choices before tools like MS Copilot, Gemini, and then a rapidly shrinking long tail of alternatives to those.
In the end what companies pay for is not tokens but results. A DIY kit of models, mac minis or whatever, and a bunch of poorly integrated OSS tools doesn't solve their problem. For the same reason, people use Office 365 rather than running Libreoffice. And for the same reason things like AWS dominate the market rather than people DIYing their infrastructure together themselves. Most of the money is in polished turn key solutions. Which is what Anthropic and OpenAI offer.
The juicy market here is the enterprise market. That's mostly business users, not programmers. They'll be hooking up all their SAAS tools (which they also over pay for), and other stuff. They'll be paying for boring things like data residency, compliance, etc. And they need access to reliable infrastructure to run all this stuff. They'll want this shit to just work and not to be dealing with a lot of poorly integrated stuff.
Most of the billions invested are being sunk into infrastructure, chip design, and access to resources (land, water, energy) needed to run data centers. A handful of companies now own most of that infrastructure and they also happen to have the top models, researchers, the best tools, and warm customer relations. And they sell access via very convenient subscriptions with high enough limits that people don't have to worry about things like token cost. The game here is recurring revenue from customers that like predictable pricing, reliable quality of service, and iron clad compliance and data security & residency, and quality guarantees. These companies don't want to be chasing model quality and have to upgrade their entire company every few weeks. They want continuity and predictability. Mostly they just pay Anthropic, OpenAI, MS, or Google to take care of this for them. There might be some niche EU players that become a bit bigger. But I don't see a large scale switching to Chinese suppliers for a full polished alternative. The Chinese might give away their models. But I don't think they'll be generating a lot of revenue.
And if you want to run your own models, you'll still need infrastructure to run it. These four companies together with the usual cloud giants control most of that and as well of the supply of resources (chips, data centers, energy, etc.) in the EU and US markets. There's going to be a long tail of self hosted and gobbled together stuff but it's going to be a much rougher experience for end users and it won't likely be most of the market any time soon.
VCs are just pass-through investors, the money comes from billionaires. And when billionaires face losing money, the whole system re-arranges itself to stop that from happening.
Yep! Maybe Chinese models are banned in USA and made completely illegal
Not sure most of money is from VCs.
Well, then let's hope you're wrong and the AI bubble won't also blow up private equity and wipe out people's retirement funds...
[dead]
I'm a civilian, not a VC. In my own case, I'm worried how many things pass through the CCP. How censorship of mentions of Tiananmen Square is something they're quite interested in
How exhausting.
3 replies →
Claude refuses to call Trump a Fascist.
2 replies →
Lets do "who's afraid of US models" version:
* Me, as an individual, because I might not be able to pay price hikes, because my revenue (salary) is much lower than what they want and I can't support my expenses via huge bank loans.
* Again, me as a new entrant to the industry, LLMs are basically pay-to-play games, again related to price hikes, new entrants might not be able to afford paying those prices 24/7 - which you need when learning new things.
* Any non-US company, US can block the models which can disrupt the whole business.
* Even some US companies, for example if you operate in EU and EU somewhat changes their mind and follow the ICC and require you to stop working with Netanyahu (war criminal as per ICC), then following laws in EU, might create trouble to your whole business.
Also me, as someone who lives in Greenland, Canada, Venezuela, Cuba, Iran, etc. China is not threatening to invade, USA is.
you live in all those places??
1 reply →
You might take a different view if you live in Taiwan.
> Any non-US company, US can block the models which can disrupt the whole business.
And read your data, see CLOUD act, PATRIOT act etc. etc.
No longer a theoretical risk in today's US political environment.
Hasn’t that been the assumption since Carnivore? Also we tapped European pols phones back in ‘12, was it? Though it’s not like the Eu are a bunch of innocents either…
It’s not a new thing. Industrial espionage has always been a thing as well. So has bribery (for deals) been a thing especially by euro concerns.
7 replies →
Also me, as someone who is living in a place that US may drop bombs on because the models may think it is a military target, or even a higher priority target like a girls school.
> Lets do "who's afraid of US models" version
Ah, cultural nuances. The title "Who's Afraid Of Chinese Models" is a riff on "Who's Afraid Of Virginia Woolf" which itself is a play on the song "Who's Afraid Of The Big Bad Wolf".
The title essentially means that the chinese models are being portrayed as the big bad wolf; but are they really the threat or are american frontier labs afraid of competition and commoditization?
It's also somewhat ironic because the author says that there is something to be feared -that western innovation will become dependent on chinese models, especially for cyber, if the american ones are restricted or unavailable.
I think your second point touches on a bigger concern for small businesses.
US labs have consistently demonstrated their intention to paywall higher intelligence. Eventually the paywall for “hyper-intelligence” will set a bar so high that average small businesses simply won’t be able to afford the bill of what is used by the top corporations to keep themselves at the top. That’s already starting, when it comes to the volumes of tokens top corporations are burning.
This is a feature of the system corporations want to establish and OAI/Anthropic are happy to oblige. 10% of a trillion dollar company is the same as 10% of 1,000,000 million dollar small businesses. Whose hitch would they rather ride, and which size customer easier to obtain to meet their revenue goal?
Not to mention, it would certainly be possible for the EU and other world powers to equally disincentivize use of US labs as a data security risk, since our top models are impossible to run in private lab environments without specialized agreements and there no access to the model weights for auditing. As best as I can tell, AI regulation is a dangerous game that is a hair away from isolationism.
In my opinion, Google is one of the few hopes in this area. There is still a paywall, but I feel like they are the closest thing to a Chinese lab we have (for frontier) in terms of their targets (real business use cases) and they actually have both the infra and already have a pipeline for small businesses into their products; they already have the wide non-AI customer base to leverage unlike Anthropic and OpenAI whose only product requires convincing people to use their (more expensive) AI.
Me as in "Unfortunately, Claude is only available in certain regions right now"
DeepSeek, Kimi, Xiaomi Mimo, Qwen, Minimax, GLM, Hy3 and Ernie are always available, and I can't be happier
in simple terms closed models are rugpull waiting to spring at unsuspecting users, it's like you take all worst components of terminal capitalism (including price cartels), subscriptions relying on almost monopolistic dependency and microsoft/uber models of hugging competition to death to remain only provider and dictate all conditions
For me, the most important factor is:
* All modern AI is a perfect front for harvesting material for processing by NSA/GCHQ.
Given the criminal US' 5-eyes/9-eyes apparatus' atrocious war crimes and human rights records, this is reason enough to eschew American AI 'products'.
I'll use the AI created by the culture that lifts a billion people out of poverty first, not that from the culture that murders children every 15 minutes and lies to itself about it ..
As someone from neither the US nor China, neither are altruistic and both have horrible histories, both meddle in my country’s affairs, and neither have my interests at heart.
China’s got plenty of blood on its hands, pretending otherwise is silly. America does, too. It’s quite easy for me to condemn both their governments and trust neither.
3 replies →
> The defining characteristic of a commodity is that it is fungible: a gallon of oil is a gallon of oil; a ton of copper is a ton of copper; a bushel of wheat is a bushel of wheat.
The concept of “commodity” as defined above is a model, a simplified abstract representation of reality, but that does not match the reality perfectly (the map != the territory).
The author claims that a token isn't literally an ideal commodity, but neither is oil or wheat, many factors influence their real value (intrinsic properties, location, available storage at production, expected delivery date, etc.) so that no two gallons of oil in different contracts have the same price.
Is treating “tokens” as a commodity a worse model than treating oil this way? It depends who you ask! I'm pretty sure that a chemist working at a refinery would be more happy to see tokens being felt with like a commodity by his company than if they started viewing crude oil like one.
(Overall, there's way too much economism in that post, and way too few facts, and as a result the argument makes very little sense, the author basically wrote that both OpenAI and Anthropic are drowning in cash right now because compute scarcity means the price must be significantly higher than the marginal cost…)
Yeah I bumped on this as well. Contra the author's claim, the analogy to energy commodities seems very direct to me. Natural gas is not useful in and of itself, what is useful is the energy or aggregates created from it, and those have very different levels of efficiency. Exactly like Sol more efficiently converting tokens into intelligence than Kimi, a combined cycle gas plant converts gas into electricity more efficiently than a simple cycle gas plant. But this does not imply that gas is not a commodity. And both the more efficient and less efficient kinds of plants have large markets; they just target different trade offs.
Edit to add: I think what he's saying is more like "tokens aren't the interesting commodity, 'intelligence' is", which makes more sense. To carry on my gas and electricity analogy, I would say the same thing about gas being the less interesting commodity than electricity, because electricity can be used for a broader set of useful things. But both things are commodities, despite one being an input and the other being an output in this case, and the conversion efficiency is one very important consideration, but not the only one.
"Intelligence" isn't a commodity at all. It doesn't make a lick of sense.
The value of intelligence is that it can solve my specific problems in ways that are satisfying to me. The example the article uses is a CRUD app - but the CRUD app I need isn't fungible with the CRUD app you need! It's not fungible at all, it's a specific solution to a specific problem that may have zero value to anyone else, and certainly cannot be replaced with anyone else's solution to their problems.
If we're comparing electricity to gasoline, then models are cars, and "intelligence" (I disagree that this is what LLMs produce, but whatever) is distance traveled.
Distance traveled is not a commodity. It's the desired outcome.
1 reply →
Releasing open weights that can approach frontier level intelligence (irrespective of number of tokens burned) is just a way of telling the world that anyone, even China, can serve frontier level inference if they have the chips and warm shells to do so.
What is stopping China from gaining a majority market share, then, in terms of serving inference?
AI Sovereignty -- yes
Cybersecurity concerns -- yes
Latency -- no, unlike previous emerging IT workload types , inference does not have strong latency requirements. eg 1s of additional network latency doesn't matter to a 15 min, 10-turn agent session.
Cost -- ultimately this comes down to a nations ability to plug chips into warm shells. which forks into geopolitical / trade on the chips side and energy scalability and modularity on the warm-shell side. Even if you call geopolitical / trade a toss-up, China has the US beat HANDILY on the energy front, yearly they are deploying 10x power to their grid relative to the US, which is shooting itself in the foot at every possible moment.
IMHO chip tech will travel across borders, absent a breakthrough in analog inference, energy scalability will ultimately dominate.
I operate an analytics site (pretty big one B2B where client's backend feeds data into our system), and we see tons of traffic originating from northwestern China (Xinjiang) from Shenzhen Tencent Computer Systems Company Limited.
There are also half a dozen other companies from China continuously hammering our clients’ websites.
I was wondering, what's in that cold dessert? Low and behold satellite imaging shows massive datacenter build outs, very cheap solar energy.
Few months ago something happened and the Geo location on data on those IP now shows "Shanghai" or "Shenzhen". A way to cover tracks? But mapping latency still points to fact that nodes behind these IPs are still operating around Xinjaing region
credit:
'You Can't Cheat Time: Finding foes and yourself with latency trilateration' https://youtu.be/_iAffzWxexA HN user: lopoc
Shenzhen vs Xinxiang is hard to do using this technique but Shanghai vs Xinxiang does show difference.
Assuming that China only distills is a huge mistake.
It’s no longer some backward place that does low value copying. Look at companies like ByteDance and Xiaomi.
Chinese companies aren’t just distilling, they’re acquiring data in the same way American companies did by paying people and crawling the internet.
The way I understand it, China has a few large companies that crawl the web at a rapid rate and build corpora. The government essentially wants select few companies to do this and then make the data available to other strategic companies operating within China.
Then there are data aggregators that buy data from apps, websites, and services, as well as systems like OpenRouter or Cursor, where companies can learn from the “traces” of coding agents, chats, and so on.
This massively reduces costs, as smaller companies like DeepSeek don’t have to do their own crawling or acquire data from 100s of websites and coding agents etc....
There are also companies in China that buy American LLM APIs and proxy them to companies within China. So, there could be 10,000+ companies using American AI products, while China logs all of this, understands how they’re being used, and trains on their traces.
And, non-state run Chinese companies are just like companies in US, they usually don't joint forces to maintain a common infrastructure, if they can build moat (or at least be in leading position for a period of time), they do it, sharing crawl dataset is no go.
I thought you got the location from BGP registration, then IMHO the before/after are both correct, it might be datacenter in Xinjiang belongs to Tencent.
What does this line mean “ the web at a rapid rate and build corpora.” , what are they trying to do? Suck up data for training AI or something else?
> It’s striking the extent to which Claude Code and Codex are proving to be quite sticky; whichever harness you start working with is likely to be the one you stick with, and that figures to be even more the case with non-technical users.
My experience has been quite the opposite. I was using Claude Code almost exclusively this winter/spring and swapped to Codex earlier this summer. It took no time whatsoever to switch. And before Claude Code, I was using Cursor. Same story.
[edit: Oh and there was also a brief interlude with Conductor, though I think they're more or less just serving the underlying Claude/Codex harness]
For personal use I agree.
For companies, these decisions are very sticky. Companies go through a lot of red tape to get anything purchased and approved, then they discourage change because it's a lot of work.
So the product that gets a foothold in a company sticks for a long time.
Then a couple years later a sales person convinces an exec that they can save some money by switching, so the switching game begins. Not necessarily motivated by the better product, mostly the price. My wife's company keeps switching their tools out from under everyone every year or two. Just when they get everything stabilized and everyone familiar with the new tool, some new contract is signed that moves them all to some other company's suite.
I'm confused, I work at a big giant Fortune 500, we all get GitHub Copilot subscriptionsn - we can switch between OpenAI and Anthropic models with just a click in Visual Studio. There's no stickiness at all. They just made us go through a training after the price hikes about how to choose between models for the best cost/benefit ratio.
9 replies →
Companies have learned their lessons on stickiness with cloud providers. Every enterprise has a multi-provider strategy now.
Every company that I've worked with that provided models internally did so through LiteLLM and offered both Anthropic and OpenAI models so it was trivial to switch between them.
5 replies →
I imagine the play here is going be connectors. Can you get slack to avoid integrating with anyone other American ai providers, same with Google suite, etc etc.
1 reply →
Same here. I flip flop between them. Most people I know who have access to both, technical or not, are doing the same. They’re just too close and sometimes one does what you want better than the other.
Have you ever worked with a non-programmer and helped them setup their AI workflows?
You install MCP connectors, specific skills, work around model/harness quirks, set security boundaries etc.
It's a lot of work, and most people will never want to change it once they have it working.
Skills are quite interoperable, and you can easily ask Codex / Claude to help you with switching the MCP connectors or any other things specific to your previous workflow. It's been quite low friction in my experience.
1 reply →
It strikes me as like setting up an IDE. People have preferences, switching is possible, but there are advantages to saying "we are a Visual Studio + Resharper shop" or "everyone uses IntelliJ to work on this project".
Yes. I taught the non-programmer to ask the harness to set up things like MCP connectors.
we have AI. WHAT is it good for if a harness cant just take a api endpoint and some permissions and duplicate.
its so distracting seeing these types of confision.
every plugin is already just multimodaling their targets.
I think they stickiness is less about the difficulty of switching and more about the lack of desire. I’ve been using Claude since day one, it works well and I’m happy, I like it. I’m sure Codex is good too. Switching from one to the other certainly isn’t going to be a game changer, the discourse shows me the differences are marginal.
Probably the only reasons I would seek change are economical.
A sticky product is one that switching away from creates a major hassle. Which means the user will pay more to avoid said hassle.
“I don’t really have a strong preference between the two” is another way of saying “the product isn’t sticky”, which is another way of saying “this provider has very little room to increase margins”
1 reply →
Convergence in coding makes them highly substitutable. But I could see harnesses configured for different purposes -- let's say, a harness for creating teaching plans -- being able to cater to its audience better than a coding harness. Maybe it's got tools to plug into standardized curricula, what the lesson books will be, what other lesson plans the district's teachers have made, etc., which could be done in a clunky way in a regular harness but could be streamlined.
agreed, my F500 company switched off claude code to copilot in 30 days. All 5k+ engineers. That is the fastest migration i've ever witnessed. This includes switching all our agents from Claude SDK to Copilot SDK.
My same progression here. I started with ChatGPT website, then Anthropic website, then Cursor, then Windsurf!, then claude, then opencode, then ohmypi, then codex, finally back on Cursor now because I think they cracked the UX for what great dev looks like. The grok 4.5 fast model + cursor ergonomics is insanely good!
The cost of me moving around these different AI models and harnesses was pretty much 0.
I use a mixture of claude code and codex and kiro as my swarm.
They communicate through my own harness, and it's working pretty well so far. claude code is being overtaken by codex however because I noticed lately the accuracy of the latter is the best.
As a Hacker News user and commenter, you are not the type of user he’s referring to.
[dead]
Which would imply that these things are fast becoming… checks notes… a commodity?
“ Let the frontier labs win by being better; don’t let them define safety or security, or pull up the ladder of humanity’s collective knowledge”
Love this.
I'm rather scared of US models - if Anthropic was the only AI provider in the world, it's easy to see that common people would have no access at all. Thankfully there is OpenAI which compete 1:1 with Anthropic (at a slightly lower cost) but most importantly the Chinese models keep Anthropic, but also OpenAI in checks.
And I'm saying this as someone working for American companies.
it's a real paradox, that chinese models are what guards democratization and private use of ai while us models are moted castles with "kings" crying that you are stealing their legally stolen goods... what times we are living in...
It blows my mind that these two companies are going to mint billionaires, meanwhile https://en.wikipedia.org/wiki/Aaron_Swartz was bullied to death by the state for sharing academic papers which were at least in part, if not largely, paid for by tax dollars.
3 replies →
ethics and other details are for humans. AI companies just proving it, even highest IQ teams are against ethics because they want more, even several millions is not enough for them.
6 replies →
The technique is called "accusation in a mirror". By accusing the enemy of doing what you are doing, when they call out what you are doing, they look like they are just weakly repeating your own accusations because they don't have any truth. And the anger that should be directed against you (because of your practices) gets directed at the enemy.
5 replies →
why is it a paradox? guarding their IP overseas has been the modus operandi of American software companies since their inception
4 replies →
Competition is a good thing for consumers.
1 reply →
The Chinese models keep American financial markets at "To the Moon" levels rather than "Igniting Jupiter as a binary star" levels. This amount of capital pressure exerts its own gravitational pull in markets and in geopolitics, and every additional dollar of valuation can propel acquisitions, which propels valuation, et cetera. Undiluted, very quickly Amazon owns countries like it today owns county governments.
this has been my tinfoil hat theory. Investors in US models might be supporting "open" models as a means to create FOMO for other investors to supply more cash to US model providers, which in turn increase their investment value. Package it so they can beat those "adversaries", equate that success with global power struggles etc etc. Seems to be quite effective.
Re: common people, yes, and given all the talk of job instability AI is creating, why hasn’t Anthropic or OpenAI offered discounted plans for laid off or displaced workers? Wouldn’t this be a tangible way to display goodwill and build adoption?
> why hasn’t Anthropic or OpenAI offered discounted plans for laid off or displaced workers?
Because it's not a widespread phenomenon. A few large tech cos laid off large swaths of people a handful of times. That's only happening in those large tech cos. Most cos are empowering their employees with AI as a tool, not a replacement, and they're not letting anyone go (unless they refuse to use this new tool).
Of course, they're not hiring as much either, since their current teams can accomplish more with AI as a tool. Maybe the AI companies could give job seekers a bit of a discount, but that would be abused to all hell without crazy administrative overhead costs, so why would they?
> why hasn’t Anthropic or OpenAI offered discounted plans for laid off or displaced workers?
because they are profit maxing. the AI world would be a completely different place if it's not for the open models.
I'm sure this will be a terribly unpopular opinion, but I've long held the view that a Chinese AI company might screw me over at some point for a complex geopolitical reason that I don't fully understand. An American AI company will screw me over tomorrow for a quick buck.
As to the argument is that (only?) the Chinese labs are training on my data, I find this almost comical given the amount of highly-personal data companies such as Meta and Google have been harvesting for decades.
I don't trust any American company because they will switch into extraction mode eventually and squeeze every cent they can out of you.
Even if the founders didn't want that, eventually the upper ranks will fill with MBAs and the board with private equity and they will make it that way. Their bonuses are based on quarterly performance not customer experience
Long gone are the companies who served their communities for decades or centuries, providing a stable return to the owners, jobs for the workers and value to the customers.
I'd rather China have my data than America. China might do something with it one day but America built exploiting it into the business model (disclaimer that I'm not Uyghur or Taiwanese though)
the article isn't saying you should be scared of chinese models
Saying GPT 5.6 Sol is 1:1 with Fable is laughable. OpenAI models are braindead in comparison and I have hundreds of hours with both.
No doubt Fable can outperform on many coding tasks, but the way you phrase it really exaggerates the gap and I'd suggest that for 95% of tasks, most devs simply don't need Fable level, and in fact, despite mostly using Claude, I have found codex is generally quicker at most tasks and can be even better than Fable with certain languages.
You have hundreds of hours with a model that was barely even released hundreds of hours ago?
The perception of capability varies greatly between task. For my needs for example sol xhigh consistently outperforms fable xhigh.
4 replies →
Try different harnesses people! I am actually preferring Chinese models at a fraction of the frontier price for coding. Yeah you need more tokens per unit of work done, but it is way cheaper still. Using CC/Opus as a staff eng / frac CTO. And Hermes/Chinese model as hopefully my team of mid levels. This way I can make good use of pro plan and then get cheap Chinese tokens for the rest and not hit a RL and know it can scale up. plus choosing your model is so cool and some are a lot less verbose.
Hermes is a better coding tool IMO. I can't put my finger on why but it just feels better. Maybe being true yolo helps.
That's nice and all, but I would not get hired in many places that are heavily regulated and risk adverse, and would not hire someone who swears by said models because there is no trust in their creators not training for malicious intent, a random tool call here and there, and you've got a "open weight model" that can send your code anywhere.
There's just no trust in a country that is digitally totalitarian and hostile towards its own people. Do people ever look at the full sized Tianamen Square photos? This is not even the photo of the many people on the ground who were killed by their government and it is still insane to look at.
https://www.reddit.com/r/pics/comments/dgua6k/the_full_tiana...
> There's just no trust in a country that is digitally totalitarian and hostile towards its own people.
Are you referring to USA, China or EU here?
9 replies →
And how is US doing on digitally totalitarian and hostile towards its own people?
In practical terms you could get US and Chinese models to review each other, right. Depends what your use case is. Coding is kinda not so bad it is reviewable and immutable/traceable per commit. An AI app that is like a psychologist or something may be more worrying.
4 replies →
Thanks, this is the first time i see these photos http://www.cnd.org/June4th/massacre.html
I think in general rest of the world needs to take notice (not saying afraid), starting with the US. It cannot be taken for granted that China's frontier labs will be a few months behind. They might be at par or exceed.
The lessons from steel, solar and EV needs to be learned by all lawmakers. You have to respect and learn from how China Government puts the system in place for complete industry takeover and they have been very good at it. The problem with AI is that democracies will be inherently slow in adopting AI, unless something changes in the system.
At minimum, every democratic Government (US, Europe, India) need to build long-term AI vision and execute that no matter which party comes to power. Additionally, be ruthless about protecting domestic labs. It can only be possible if the intelligence pricing by domestic labs per productive task is in the similar range as open-weights models. Right now, it is not the case, even if the article gives the example of Sol vs K3.
Protecting domestic labs means not bailout, but fast track to cheapest energy, fast track approval for data centers, enforce some guardrails so customers get to use the open weights models only hosted in the country by US (or Europe) businesses. Without these protections, it might be a slow death.
In the earlier days of the USA we did the same thing, with our government having an industrial policy that fed US industry and put us ahead of Great Britain.
It doesn't have anything to do with the form of government, it has to do with the aims of the government.
What is this panic and protectionism supposed to be good for? This is open source software, there is no "AI industry", there's virtually nobody employed in this. "Domestic AI" makes about as much sense as a domestic Linux kernel. If the Chinese want to subsidize the world's water and energy use to supply the world with chatbots good luck to them. There's no need for guardrails or fast track data centers, they can plaster their entire country with data centers to churn out slop, I'm glad we don't
I think it is deeper than that. "LLM => peak of productivity" takes way less time and effort than "Linux kernel => any productive work". Compare Dec 2025 vs July 2026 models in terms of capabilities.
Nobody can predict 5 year out. However, the country that can be ultra efficient by making their governance, health, manufacturing, military, etc AI-native will be far ahead in the game.
1 reply →
Only those who intended to use the new technology as a tool to achieve dominance and control are freaking out.
And it's the only viable tool the US has left. It's reasonable they are freaking out.
Ok…
Excellent article; the argument towards the end for allowing distillation for US companies is compelling:
> To that end, here’s an even more interesting question around distillation: why exactly is it bad? After all, what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models that are themselves being distilled? Who is exactly being wronged here?
> In fact, this paradox is the solution. I believe that open weight models are good for innovation (and, per the above, I think that labs on the frontier will be fine), but it’s a problem to be dependent on China. The U.S. should pass a law that (1) makes explicit that collecting data for training models is fair use, and (2) bars terms of service that forbid distillation, for U.S. companies at a minimum. Stopping distillation — which is literally just querying the API — is nearly impossible; the U.S. should go the other way and lean into a new copyright policy that both indemnifies the labs and also guarantees that what they learned fuels further innovation for everyone else.
Very good point.
That would prevent the facebook strategy of sucking up MySpace users and then defending TOS that prevent other social media apps from doing the same to them.
The U.S. "executive" class is so obsessed with the "exploit" part of the explore/exploit cycle that it's very clear they are prematurely closing advancement. Better a little money and power for them now than a lot of money and power for their country/humanity.
This has an element of stochastic improvement so it's hard to predict but the chance of the U.S. "winning" this "race" is pretty bleak.
You see this all the time in communities that have internalized hierarchy as a "good", little kings of shit mountain vying for less and less at a higher and higher cost.
My personal hypothesis here is the Chinese government looked at the game and simply decided not to play:
An astute Chinese analyst could reasonably forecast that they had little chance of controlling the AI market due to sovereign trust issues, but would also note that AIs are just software.
When the dust settles the US still won't have factories, and the real value of AI models is still going to be embodying them and getting them to do real, consumer facing work.
Perhaps the most striking thing about the AI boom is how quickly the US abandoned the veneer of local manufacturing in favor of more expensive buildings producing nothing you couldn't make anywhere else on the planet...from imported parts.
Yeah, how much of this is China waving distracting AI hands over here while the US Genius-In-Charge watches and completely ignores reality.
According to openAI's own @deanwball: Even OpenAI isn't buying this distillation talk:
https://xcancel.com/deanwball/status/2078133895766114412#m
> open models are inherently decelerationist
I’m struggling to understand this perspective. Is he using the words accelerationist/decelerationist in a sense other than the obvious one?
EDIT: I searched his twitter history and discovered that his argument is basically “if you drive down costs, then OpenAI will have less money to invest in development, slowing down the overall rate of AI progress.” IMO this take betrays an overwhelmingly stupid degree of exceptionalism, but I guess that’s what I’d expect from someone working at OpenAI.
Nah, the point is that if models are commoditized and there's no hope of making significant profits, no one is going to be willing to make the massive investments necessary to continue pushing the scale frontier. How large a training run do you expect investors to fund out of the goodness of their hearts?
4 replies →
What a delusional f-wit lol he wants protection of profits for reinvestment?
Every company wants that!
This guy is predicting AI covid escaping from a Chinese lab. I find that kind of silly
here's the original X link : https://x.com/deanwball/status/2078133895766114412#m
So he thinks open weight models will lead to “AI communism” and “dystopian hell” and in the very next point proposes that the US create a federal agency to discourage the use of open weight Chinese models. The motivated reasoning in this post is unreal.
Can you or someone please explain several of the claims made in this tweet?
"I am personally surprised the Chinese state continues to allow the open sourcing of models this good, given potential risks" what risks?
I suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI). Confused what this means
Open-weight models are inherently decelerationist, and I'm continually surprised to see the so-called "accelerationists" so excited about open-weight models. Confused again.
One probable outcome of an open-weight-model-dominant world is full AI communism, which is precisely what China proposes: rather than a market product, AI is a "public good" which will ultimately be provided by the state as a kind of "digital public infrastructure." This future strikes me as a dystopian hellscape, but I've never met an open-weight models advocate who doesn't ultimately concede this is where things end. I don't understand this at all.
Can someone in the know please use plain layman's terms to explain what this tweet is about?
> I suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI). Confused what this means
I think it's referring to the belief that LLMs are not the path towards AGI, and that LLM's, while useful, are not going to have the impact that the American labs believe it will have.
>Can someone in the know please use plain layman's terms to explain what this tweet is about?
The Silicon Valley people like this openai guy, high on their own supply, are convinced they are building some machine god that will either bring about the end of the human race or utopia, they therefore cannot understand why the Chinese (or any other normal person on earth) are not afraid of chatbots and have other things on their minds.
2 replies →
> I am personally surprised the Chinese state continues to allow the open sourcing of models this good
Writer seems to have no clue how IP actually functions in China
> "I am personally surprised the Chinese state continues to allow the open sourcing of models this good, given potential risks" what risks?
I assume they mean the risk of opening up "forbidden" knowledge to the masses without adequate control, which the CCP hasn't historically been known to do.
> I suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI). Confused what this means
Yann Lecun is a pioneer in the field of AI and Meta's former AI head. He is famously anti-LLM, and considers the entire technology a dead end to achieving human-level AI. The author is saying the CCP has similar views (that LLMs aren't going to get exponentially better/lead to AGI) which is leading them to not control these models as tightly as they otherwise would.
> Open-weight models are inherently decelerationist, and I'm continually surprised to see the so-called "accelerationists" so excited about open-weight models. Confused again.
"AI accelerationists" = people who want AI to progress. According to the author these people should not celebrate open models because open source = less commerical value in LLMs = less investment into the field (because how are companies going to get returns?), and this will ultimately lead to slower growth.
The last bit is about government controlling AI vs commercial companies. According to the author the former is a dystopian hellscape.
IMO even if you think his points make sense, his job title ("head of strategic futures @openai") means they should all be taken with a massive grain of salt.
1 reply →
Chinese ai is bad. It’s slowing down progress and it’s so bad we called out the c word and asked for more regulation. Basically advocating for more government assistance to openai
> This is a point that bears repeating: because U.S. open weight model makers must follow the frontier labs’ terms of service, they (1) are worse than Chinese alternatives and (2) end up distilling the distillation, just with a detour through Chinese labs. Wouldn’t it be better if western open weight model makers could go to the source?
This is of course a baseless assumption. Let's say China created GPT 3.5. Then I can guarantee you that Ben would say "Western frontier labs are at a disadvantage when gathering data, because they have to follow the terms of service of Western media, and Western copyright law". Which we now know wasn't true.
And sure, some will say "but Anthropic can more easily block this as it's a single point of failure". But it's doable to overcome this. Without being "state backed".
One thing I have not seen mentioned between Chinese AI vs US, population.
China has a billion+ people that their AI can "study". Plus due to China's political structure, their AI has access to everyone's chats, comments and sites, scraping everyting.
Here in the US, with 1/3 the population, the AI race was lost before it even began. Plus in the US, all companies and people are doing all they can to restrict AI from scraping sites and peoples chats.
So I believe, China will end up owing AI.
"China will end up owing (sic) AI"
I think you hit the nail on the head - right there!
People who claim that the Chinese open weight models have some type of manifest advantage don't realize that the close weight models have a huge advantage as well: the researchers from OpenAI, Anthropic, Google, xAI, Meta are not dumb, they can read the white papers written by DeepSeek, Moonshot, etc, and they can inspect all those architectures and they can pick and choose the best tricks there are out there, and of course, they have access to their own in-house secret sauces.
Sure, any model that is not at the frontier can use the frontier model to generate synthetic high quality training data, so this can reduce significantly the training costs.
But at the scale of OpenAI, Anthropic and Google, it is quite likely that the (raw) training cost is very high anymore. Here's a few heuristics:
1. All the hyperscalers see a huge demand for inference. They can't deploy datacenters quickly enough to satiate all the demand they see. But, it's is impossible for the inference demand to be constant throughout a day or a week. If you use the times when the demand is lower than the peak demand (which is almost all the time) to dedicate the spare compute capacity to training, then your the cost of training compute is zero.
2. It is likely that increasingly a higher cost of the "training" is actually setting the guardrails, which is essentially post-training. As we've seen, without proper guardrails, the US Government won't allow you to serve inference. Anthropic was hit directly, but OpenAI delayed their 5.6 release as well to make sure the US Government is ok. This part of the training cost can't be reduced easily by using synthetic data generated by other models.
3. The frontier labs are also investing more and more in building an ecosystem around their models.
I am not a frontier lab insider, but take a look at the jobs posted on the Anthropic career page [1]. There are 74 jobs in "AI Research and Engineering" and by my count at most 15-20 are related to pure model training (of pre-training or RL type), and the rest are post-training, safety and security, alignment, interpretability, productivity and lots and lots of other things.
[1] https://www.anthropic.com/careers/jobs
People who claim that Postgres has some type of manifest advantage don't realize that Oracle has a huge advantage as well…etc
If the Google and meta engineers are not dumb how come they consistently trail behind the frontier labs and even the Chinese labs with a fraction of the funding.
Probably bad leadership
I always suspect they have the most to lose if legal decisions on copyright issues don't go their way.
Imagine a scenario (theoretically possible but increasingly unlikely) where a US court decides that using "pirated" copyright data to train models is illegal. Now the AI developer has invested hundreds of billions of capital into a thing that is declared illegal and has to be scrapped.
This risk affects existing megacorps more than "startups" like OpenAI and Anthropic (and Chinese companies), because the megacorps have much more to lose. They actually have the cash to pay damages if the flood of copyright claims arrive at the door. This will not only bomb their AI development, but also the rest of their established businesses as well.
And thus I strongly suspect legal issues are holding them back a bit. Megacorps want to win the AI race, but not to the extent they stake the rest of their established business, while the newer companies' only product is AI, so they have to go all in.
Notice for example how Meta's Llama performed much more poorly after they got smacked by a bunch of lawsuits claiming that they torrented a bunch of copyright data.
(Disclaimer: I'm an outsider and everything I base my speculations on is public knowledge.)
That plus they don’t distill so they have worse RL examples.
2 replies →
they, rightfully so, have no faith in LLMs.
>they have access to their own in-house secret sauces.
I remember some feature lauded by Gemini was reverse engineered by the open weights guys in < 30 days.
If they dont publish some technical information its hard to protect in the US, but conversely, once it is published smart people from outside the copyrightosphere can start working to reverse engineer it.
>3. The frontier labs are also investing more and more in building an ecosystem around their models.
Theres nothing there that isnt immediately replaceable.
> There's nothing there that isn't immediately replaceable.
Indeed. But that was not my point. My point is that we still have this old impression that training cost is dominated by compute and it is hugely expensive, and the Chinese labs can short circuit that by distilling the American frontier models. I don't think the training compute cost is a big factor anymore for the American frontier models, because of the reasons I gave. If the Chinese models can get the training compute cost down by a factor of 100, that's not going to make them 100 times cheaper, and not even cheaper by a factor of 2. Maybe 10% cheaper or so.
It’s giving desperate!
"distillation attack" is such a loaded term that really pisses me off.
Distillation is a technical term with real meaning, and historically requires logits which Anthropic does not provide.
"Generated training data" is the correct term. It's not an "attack". And Anthropic undoubtedly also generates training data for each new generation of models, yet you never see them claim Fable is a distilled Opus.
1) Model distillation is the process of transferring knowledge from a large model to a smaller one. It doesn't require logits. https://en.wikipedia.org/wiki/Knowledge_distillation
2) The word "attack" is standard security vocabulary. Per RFC 4949:
There are hundreds of named "attacks".
3) The "attack" part of "distillation attack" refers to distillers creating tens of thousands of fraudulent accounts, using proxies to bypass georestrictions, deepfaked IDs, and paying real people to pass biometric KYC checks. Who then blended this in with real user traffic to conceal their behavior.
It doesn't refer to the AI training technique in any way.
If they acquired this data without the fraud, you'd have a point.
Sure that can be called an attack, but then we must also concede these labs essentially massively attacked everyone else in existence to get the data, and continue attacking as we speak.
In a way you could see this as a case of Robin Hood. The US companies exfiltrated all the data on the planet just to hoard it for themselves now and accuse anyone who tries to get a piece of that back from them, and the Chinese labs are distilling it to offer it for cheap.
Obviously a bit more complicated than that but it still holds pretty well.
> Model distillation is the process of transferring knowledge from a large model to a smaller one
Sure, and large-to-small is a key part of the definition, and why it's called distillation (cf concentrating something). When Anthopic use synthetic data generated by Opus to train Sonnet or Haiku, then this can correctly be considered a type of distillation.
When Anthropic accuse Chinese companies of "distillation", it seems they are using this word to refer to two potential uses of their model outputs:
1) Using Anthropic model outputs (aka synthetic data) as training data, especially for reasoning, for Chinese models. This really isn't distillation though, since (unlike when they distill their own models) Anthropic don't actually provide the reasoning in their model output, only a "summary" designed to hide the actual reasoning. You can't distill what you are not given!
2) Another way Chinese companies may be using US LLMs is for "LLM as judge" where you are just asking the model to use it's expertise to judge/rate something that you provided yourself (to provide RL training rewards), although for coding you really want hard rewards which are easy to obtain, not fuzzy "looks good to me" ones.
Of course Anthropic are trying to pull the drawbridge up after themselves and their TOS says you can't use their models to develop anything that competes with them, and this seems to be what they are generically referring to as "distillation" - any use of their models that they suspect is being used by the Chinese to improve their own models, not just what what might more technically be called distillation, unless you want to define that word so broadly that it does mean this!
1 reply →
I'm not even convinced that this fits your definition. A distillation "attack" doesn't evade the security system in the sense of hacking past a login. The only part of the "security system" that it bypasses is the Terms of Service. And that's only after said data was acquired legally and correctly and normally.
It's a post-facto attack, which doesn't sit right linguistically to me.
1 reply →
1) Your own Wikipedia link goes on to describe using logits. Yes, language evolves to mean multiple things, and that is my point: Anthropic is pushing for a watered down definition. Furthermore, Anthropic hides thinking, so you do not even really get model outputs, you get some downstream partials. Further-furthermore, Anthropic does not describe their own models as distilled when they produce training data. Why? Because generating training data != distillation.
2) This is a stretch: it allows Anthropic to arbitrarily define "attack" via TOS, and ignores the fact that the generated training data is literally paid for by the "attackers".
So what kind of attack are AI companies doing by scraping up copyrighted info to build these LLMs?
"You're trying to kidnap what I've rightfully stolen."
> 3) The "attack" part of "distillation attack" refers to distillers creating tens of thousands of fraudulent accounts, using proxies to bypass georestrictions, deepfaked IDs, and paying real people to pass biometric KYC checks. Who then blended this in with real user traffic to conceal their behavior.
Lol. Isn't this literally many of the same tactics OpenAI and Anthropic used to scrape the internet? So now it's an "attack", but previously it was just "training".
1 reply →
Completely unrelated, but I'm seeing people and especially LLMs using causal/intervention so much it's kind of driving me insane.
It's actually a very goated term but not everything is causal, it also has precise technical meanings (although those get blurred too given that causal can mean anything from intervention proper, to mere depdnence on something prior)
Are you talking about my username? Yeah I liked the word before LLMs made it cool/uncool.
I like this article. Rings very true.
I like Anthropic, I don't think all their talk of safety is bluff and bluster, or at least, I want to believe that the people who left OpenAI because it had lost its focus of helping humanity still want that to be their main goal. However, yes, it seems that business fears are once again causing those in charge to turn "we want to help humanity" into "we are the only ones who can help humanity, and therefore we need to be the most profitable, and the only survivors".
If you want the former ideal to survive, at Anthropic and outside of it, you need to be willing to collaborate beyond profit incentives and recouping capex. Show other labs a commitment to research and community and they will follow. Better to bring teams together rather than implicitly say you distrust them, pushing them that way instead.
"Anthropic and OpenAI likely have among the lowest costs per unit of frontier-quality intelligence"
That's a big claim that his whole thesis rests on but is largely not backed up. Where are the apples-to-apples tokens-to-answer benchmarks that he's using - doesn't look like there are any, just a handwavy implication that US models are more token efficient, which they may be. But how is there so little effort in establishing this point in the article? And US labs may be in much different situations from one another: it's known that some labs like OpenAI bought big, early on compute and may have secured better pricing.
His article also does not mention the average price of electricity in China vs the US, which it seems like China leads on, and probably has the political power to more heavily subsidize. While I agree the COGS is often overlooked by top line benchmarks on coding tasks, etc, it seems that he's running on a big assumption while claiming "labs on the frontier will be fine".
But.. if you are running Chinese model in the US, what difference does it make? Isn't the whole "scare" (khm khm) with Kimis is that now I don't need Claude, cause I can run Kimi on my own hardware in my own datacenter and it's maybe not as good as Claude July edition but it's is as good as Claude January edition.
It doesn't need to be as good. You can route to the appropriate model and save so much money.
I have sonnet do the thinking, deepseek does all the tasks. I've massively reduced costs with this approach.
Sure, and I think that flexibility further undercuts his "frontier labs will be fine" take, which depends on top US labs having pricing power.
What's wrong if the roles of USA and China are reversed in technology? Why does the rest of the world care? It's not as if USA has done a great good for the world, and China has evil intentions towards the world. Infact it is the opposite in the case of AI so far.
I feel like I'm crazy here but isn't China just sitting there doing nothing?
On one hand there's the relentless barage of American propaganda. I get that a militaristic society needs an enemy to fight against lest they turn on each other. I get that if you tell people a bad guy is coming for your jobs or your lives then you can maybe get your workers to accept worse conditions and living standards which increases profits. It's ghoulish but there's some logic.
On the other hand I can't see China doing anything except minding its own business. No tariffs. No bullying other nations. No wars started. No threatening allies. They don't let their people waste their lives on brain rot or gambling. They largely align with UN resolutions. They respect international institutions instead of always being the asterisk.
Has American propaganda just failed to work outside it's borders? It's not landing at all
Datapoint from the cheap end of the market: I run local models on a couple of Orange Pi SBCs and a decade-old Optiplex with no GPU. What runs usably on that class of hardware is almost entirely Chinese open weights — Qwen's MoE builds (35B total, ~3B active) are the only thing that gives me acceptable speed on CPU, with Gemma as about the only western exception. I evaluated Kimi too and ruled it out purely on size.
Whatever the strategic picture is at the top, at the bottom of the market "weights you can download and run on hardware you already own" is the whole ballgame, and right now that's mostly Alibaba's to lose.
I assume this is slow and that makes me curious - what are you using this for?
[flagged]
Is Gemma worse?
yes. less stable. more hallucinations
[flagged]
I think releasing models for free is some 4d chess move by the chinese. Big chunk of the us stock market is fueled by ai mania, if the frontier labs turn out to be drastically less valuable than first believed, the downturm may be very bad. Think of all the big tech companies that have a ton of debt that they took to pour money into AI. It seems like a similar tactic to what Chinese car manufacturers are doing in Europe but the result may be more dramatic.
I loved this article! Regardless of how you feel about AI as an industry or tool, the economics of AI is fascinating. It's awesome to see something like this that gets into the business side a bit more.
I don't know if I agreed totally with the assessment of the risk Chinese labs pose to US labs though, in particular I think the main part I wasn't sure about was this:
> I highly doubt that Chinese models are cheaper to serve on a marginal cost basis, they just seem cheaper because Anthropic and OpenAI are so supply constrained that they are charging far more than they would if there were sufficient supply to meet the demand for intelligence.
How true is this? My understanding from Deepseek's original paper was that they focused heavily on optimising training and inference costs, in particular so that they can operate on cheaper (and more accessible to China) hardware.
It's possible I'm just not in the loop, but nobody seems to talk about US models innovating in this way (I'm just talking about cost-to-serve/train, not saying US AI companies don't innovate in other ways).
It seems to me at least, like there's a fair bit of evidence that AI shifting to a price based commodity market (vs a "best-model takes all" type market) would put China at a significant advantage? And even more significantly, require a pretty hefty correction of company valuations in the US?
He also forgot Europe in the equation. More and more companies use Chinese open models on European inference because of geopolitical concerns and data privacy, which could be a problem for the big US ai labs if they loose on the market.
I am more afraid of the US models to be fair. A country with no clear direction in many regards , that is threatening day in and day out the rest of the World for its own interests.
I fully agree with everything in this essay. Make distillation fair use. And let us use Mythos/Fable and Sol and successor or future models for all cybersecurity purposes.
Maybe not the main point of the article, but I have a doubt about the author's introduction to commoditized markets:
> - Supplier A will sell 10 units of the commodity for $20, earning $10/unit
> - Supplier B will sell 10 units of the commodity for $20, earning $5/unit
> - Supplier C will sell 5 units of the commodity for $20, earning $0/unit
> ...
> Bankruptcy risk is where fixed costs come back to the forefront: Supplier C has both fixed costs (like potentially R&D spend) and also may have taken on debt [...] It can’t price its commodity with these costs in mind — remember, the market-clearing price approximates the marginal cost of the highest-cost unit needed to satisfy demand [...]
Why can't Supplier C price their fixed costs and debt into their product? The entire reason Suppliers A and B are earning $10 and $5 per unit, and not more, is because they cannot meet demand by themselves and are therefore at the mercy of how much Supplier C is willing to charge. Couldn't Supplier C just refuse to offer 5 units of the product at a price that would bankrupt them?
Sincerely, an interested observer of business/economics.
I disagree with the first half quite a bit, COGS ultimately depends on the use case. If someone just wants something that a smaller model can do, running a local model on phone is going to have a negligible cost close to running any other piece of software. The alternatives to running a model also determines COGS, and even Jensen Huang has distinguished between the job and the work for the job that AI is capable of doing. Smaller models are always going to win in efficiency too.
I also heavily disagree with this no-marginal cost in software distribution view whenever I see it, bit rot is real, and someone is paying a marginal cost whenever they do an update. You have to re-distribute with changes whenever anything changes. These costs are just hidden because things are ad-supported or bundled in some way. These costs are also kept low because of standards and open source, but could become high anytime. Additional licensing also has costs.
That said, I couldn't agree more with the last paragraph, charging a high price for models would be better than denying access for any model that wants to stay relevant.
While intelligence is said to be a replaceable commodity, oil and copper can be used in nearly the same way even if you change suppliers as long as the quality grade is matched. However, I question whether two models that produce the same benchmark answers are actually interchangeable in real world use.
Personally, I think models will increasingly become specialized in different areas, some good at X, others good at Y, and we might see workflows that mix multiple models.
The article makes a great point that the token industry is going to be commoditized as time goes on.
Following this argument the key for each player will be the underlying cost structure and serving capacity to offset the upfront R&D cost.
The cost infrastructure will be driven by access to cheap electricity and cheap chips. The capacity will be driven primarily by depth of pockets now to buy all available supply in chips/mem/data center building capacity. While China is certainly in the lead on cheap energy, I am wondering if they can/want to beat the > 1tn USD being spent on data centers right now. Following the example in the article:
If company C from China sells 10 units for 20 USD produced for 10 USD they pocket 100 USD.
If company A from America can sell 100 units for 20 USD produced for 15 units, they pocket 500 USD or 5/6th of the market's profits.
> By the same token, don’t expect China to do anything about distillation attacks on the frontier labs. I think it is mistaken to attribute all of the success of Chinese labs to distillation, but it’s just as much of a mistake to pretend like distillation doesn’t give Chinese labs a big advantage.
I think we see this with Meta being paranoid about internal Claude usage, to avoid inadvertently distilling[1].
If distillation is a driver, then smaller American labs could be distilling, but are not for legal reasons.
But that's a big if we just don't know for sure.
1 - https://cryptobriefing.com/meta-restricts-claude-code-codex-...
I'm worried that any ban on Chinese AI models might be an excuse to get mass surveillance.
You don't need mass surveillance to enforce such a ban. Once the US Govt declares Chinese AI models are banned, no US business will use them nor distribute them. Any cloud service that rents out GPUs in the USA will explicitly prohibit the use of Chinese open model weights in their terms of service (you open yourself to a lawsuit if you violate their ToS). Any Tokens-as-a-Service provider will refuse to serve those tokens to customers in the US.
Sure as an indie hacker, you could go download the weights for a Chinese model with a VPN, and then attempt to run it at home by building your own GPU cluster but these large models require quite expensive hardware to run on and so it makes it less likely than anyone would invest that much capital to do something that is illegal. There's no way for them to sell a legal service using those tokens. So it can only be strictly for personal use (the Govt won't care because very few people will have that kind of money and risk appetite). The other option will be that there will be some shady third-party providers in foreign countries who are willing to sell tokens from these models to US consumers knowingly.
> Any Tokens-as-a-Service provider will refuse to serve those tokens to customers in the US.
So under a ban rest-of-world gets to use cheap open-weight models but American companies/individuals must only use only ‘approved models from US for-profits’? Doesn’t seem like that kind of protectionism will be popular or politically tenable. Not so long ago US chose cheap TVs over maintaining the country’s manufacturing base.
(Despite what you wrote it’s also really hard to imagine that enforcement wouldn’t leak like a sieve. Unser sufficient economic incentives [which are the predicate for the ban], loopholes will be found.)
You couldn't be more wrong. What's being sold are chunks of time with access to specialized hardware resources. Through which model or with which device is irrelevant; the one winning, and continuing to win for some time, is Nvidia, and that company is American. As long as they have the H100, B200, or B300, China won't be able to compete with the American strategy, no matter how many new models they release, because these types of cards require incredibly powerful hardware to run.
No one should be afraid of anything. Fear is a terrible advisor. Keep your eyes open, try to read the context as careful as you can and adapt as best as you can. Don’t spent too much time trying to be an oracle, never works out…
Something I did not realize. The White House instructed intelligence agencies to help in preventing distillation of US models.
Whether or not distillation matters a small amount or a big amount, still interesting:
https://www.whitehouse.gov/presidential-actions/2026/06/nati...
"Right now, none of the above analysis applies because demand exceeds supply for frontier models, and supply is limited by a lack of compute."
It gets particularly hairy because models themselves can tune their "token verbosity" to manufacture demand for compute. If compute was such a precious resource, you'd think we'd be complaining that the output was too terse.
The ability for a vendor to determine ex post facto how much a query costs is a similarly new economic phenomenon to zero marginal cost.
The author doesn't seem to realize that a healthy margin has been built into the inference pricing. Once low cost open source inference providers get their hands on powerful frontier level models, there would be a severe margin compression for OpenAI and Anthropic.
Source: https://martinalderson.com/posts/the-upcoming-ai-margin-coll...
Why do you think an inference provider competing for the same compute as OpenAI and Anthropic gives up those margins rather than giving a modest discount over the frontier for near-frontier performance?
> I expect the inference market to grow much faster than training costs
This was my assumption as well. It's also generally true of 'traditional' deep learning models that inference cost is expensive compared to training.
But the cost per token for inference has been very quickly dropping. I don't recall where, but I recall about ~50x down from GPT3, even as model complexity has increased. Even with agentic systems, there are lots of optimization opportunities. I'm less assured about claims like this.
> [Anthropic/OpenAI] are serving models at a particular capability level for months before their competitors, and are simultaneously applying the best models to optimizing those costs. Second, intelligence isn’t in fact a perfect commodity, in part because applied intelligence makes itself smarter
Is he casually assuming a singularity has already happened? A regular first-mover advantage I can understand, but those have been squandered or lost many times before.
No, he was just talking about reinforcement learning, etc.
So that's just an appeal to authority (longevity?) not anchored in reality. Just because you've been doing something for longer doesn't mean you're the best at it; Google has been shipping AI/ML models years before the founding of OpenAI and Anthropic, but it's playing catch-up on LLMs.
Haven’t we been in this “China is 3-6 months behind” for a while now (maybe up to a year? Longer?)
The actual difference is how much scrutiny and time was put into the Mythos / Fable and GPT 5.6 release. Making it feel like “these are a big deal”. Spring and summer THAT was the AI story
Then Chinese labs release models that approach Fable performance. We’re shocked they just seemed to appear out of nowhere.
It’s less about the gap closing. It’s more about the weight we put into Fable-capable models.
Me I am, so very afraid of actually decently priced inference.
Just couple years ago mr altman was promising open AI, to benefit humanity. They even forgot to remove this part about "openess" from the company name.
Today chineese deliver that promise and usa people freak out like they have any skin in this game. Enjoy the ride leader of the free world....
I still can’t wrap my head around how transferring all assets from a non-profit foundation to a for-profit company was legal.
altman should listen to his own words, and he should replace himself (wasting energy for eating) with ai (supposedly not wasting energy for eating)
it would be beneficial both for openai and world (most likely)
It really is amazing that China went from the country that hacked Google out of its market to a trusted source of AI in the tech world.
Distilling models using more advanced LLM is not a new phenomena. It is a cost effective strategy in a highly competitive emerging field.
Also, the Hidden-Agent problem exists in every model, and is a persistent tangible risk independent of whatever team people cheer for at the games. Let us remember, every LLM nuked all of humanity 92% of the time in simulated war games. =3
Partial summary: By giving away weights for free, China can drive the price of white collar labor towards zero which is what the USA economy is built on. China which is built on manufacturing will not be as affected.
I'm afraid of Chinese models because they are rip-offs of other models...and there's no telling what your data is being used for when you utilize the API. It's one thing to have US companies using my data and another to have a Chinese company who aren't bound by any IP laws jacking all of my codebase.
IP laws work if the penalty is larger than the use case of the data. Pretty sure literally every US company is using the data knowing they won't be punished equivalently. We're living in the age of trillionaires, Facebook paid 5b$ for the cambridge analytica scandal, something Elmo could write off as a business expense at this point...
Open weights dude. You can literally run it on your own or rented hardware and give your data to exactly nobody, unlike closed models.
A reminder any comment about risk FROM china, invites a "Tu Qoque" facing the other way. The paranoia here is probably fully symmetrical.
I see massive risks in belief the inferences drawn from strategic information cannot be seen. So if you depend on some position remaining inside a secure facility but you drove to it from data outside that secure facilty, The likelihood that an inference model can derive the same idea is very high. Collation over public data is not inherently secret because you used a secret model or secret weights.
A more simplistic take might be that the fear is not actually driven in the secrets, the fear is "the emperor has no clothes"
I'm afraid to trust technology coming from companies operating under the jurisdiction of a rogue aggressive nation that is continuously attacking other nations both (so called) ally and foe using economic and military actions.
Sir, given what's happening in the world these days, I cannot really tell which nation you meant by your statement :D
Really? Ok, name list of rogue aggressive nations that are continuously attacking other nations both allies and foes, and I tell you which one he meant...
3 replies →
Well done, sir. This post has evoked the expected responses from the other comments.
Are you describing the United States or China with this quote? It's hard to tell.
Outside of a few boarder disputes with India, I don't China has militarily attacked anyone since they got their ass handed to them by Vietnam (Sino-Vietnamese War 1979). So I think that rules out China.
4 replies →
Are you talking about US companies, right?
your loss ::shrug::
Not saying the heavy hand of the Xi admin and Chinese communism is equivalent to the outright lawless, corrupt, grifting and vindictive and hateful administration that is Trump 2.0 ... but it's pretty much * makes the 6 7 motion that the kids do * this for me.
China will just do a better job -- if they do this at all -- of storing, collating, indexing, and using data from their state sponsored and championed AI labs to use that against the US. [1]
The US under this admin is doing the same, attacking universities, allies, it's own citizens.
The two governments are operating more or less the same. Ergo the ai models from each country's ai companies shant be trusted either.
[1] https://www.abc.net.au/news/2019-05-16/grindr-why-is-the-us-... The late Lindsay Graham comes to mind. Even larger bombshells probably.
I think chinese componies are not doing charity for releasing their models publicly. That is the strategically best decision they can do for now. Morally they should but IMHO they are not angels :D.
Why hasnt the EU develop an competitive open source model? Ive only of mistral, but with the chinese models you have GLM, Kimi, Qwen, Deepseek etc all of which seems to be better and better
Because the European researchers and engineers are busy building models in the US
Is there more info?
New, smaller models can outperform the previous generation's foundation models.
What if there's a way to extract the commodity of intelligence from smaller models?
I've seen for many use cases it's well enough. :)
But it’s a problem to be dependent on China. The U.S. should pass a law that (1) makes explicit that collecting data for training models is fair use, and (2) bars terms of service that forbid distillation, for U.S. companies at a minimum.
I'm amazed that no one is talking about proposals that are surely being discussed in Washington and pushed by SV lobbyists to restrict Chinese models on national security grounds, or other some other basis.
The belief that Bytedance could engineer a finger on the algorithmic scales to serve the interests of the Chinese Communist Party led to a lot of debate in Washington, and ultimately resulted in TikTok being divested from its Chinese owners. Huawei is shut out from the U.S. market, which limits its business even in markets where it's not banned because it's effectively stamped with a scarlet letter.
IMHO, Chinese models are headed for a similar fate or at least a showdown in Washington or the courts because they are supported and/or controlled by entities which ultimately serve the CCP.
So happy that we have finally 2 countries playing the competitive game. No more secret deals between competitors. No real competition. A race to the bottom is always a good thing for consumers.
My thinking is that with the current narratives out of washington we are on track for a ban on Chinese models and possibly sanctions against Chinese AI companies
I think it is the right move to protect American interests
Related:
Ben Thompson is wrong: US frontier labs are right to be panicking
https://news.ycombinator.com/item?id=48982061
the leader of China praised Open Source in a speech. Crazy times
> because U.S. open weight model makers must follow the frontier labs’ terms of service, they (1) are worse than Chinese alternatives and (2) end up distilling the distillation, just with a detour through Chinese labs. Wouldn’t it be better if western open weight model makers could go to the source?
Is this an assertion that is backed by evidence?
From the Elon/OpenAI trial:
> On the stand in a California federal court on Thursday, Elon Musk was asked if xAI has used distillation techniques on OpenAI models to train Grok, and he asserted it was a general practice among AI companies. Asked if that meant “yes,” he said, “Partly.”
https://techcrunch.com/2026/04/30/elon-musk-testifies-that-x...
The best model is the model that runs best on your hardware.
They are just afraid open source/weights models, and the models are from China.
If EU build some SOTA open source models, they will design a different story
US restrictions are excessive, and companies have to protect their infrastructure with the same technology they are trying to block.
I think a simple experiment is enough to understand why one would have some concern with a state-censored AI model : Just try asking them about atrocities committed by their state[1].
[1] : https://imgur.com/gallery/ai-models-on-atrocities-B7DKUXc
It's hard to overstate how important this point is. If China is the standard of openness, it's a pretty low standard. Your point I think adds to what the article is saying, from a different perspective, but arrives at a similar place. Our best selves in many ways are defined by openness, unflinching self reflection, and competition. We should remember that, and as the author of the article says, lean into it rather than letting fear mongering and histrionics protect these models from competition.
Yes and no? I've had the same experience with asking about Tienanmen square -but then when asking a Deepseek v4 model (and confirmed by asking the chat on the deepseek site) about the "laying flat" movement it gave me a detailed answer that was unexpectedly sympathetic to the movement.
ChatGPT supports left wing American narrative and according to them Vietnam war was American's fault. Could you try asking people killed due to left wing American ideology.
ChatGPT tends to support any narrative that it thinks the user supports.
I just asked ChatGPT about the Vietnam war and it did not say that it was purely the US's fault: https://imgur.com/a/zmiOyuu
It also didn't seem to have a problem describing people killed for left wing ideology: https://imgur.com/a/LhH9saL
These are both with the free ChatGPT membership, as I do not have a paid membership anymore.
I know this is a common trope to bitch about, but honest question: did you actually try this before you commented?
ETA:
I was curious what something that was trained around me specifically (fairly typical lefty American progressive) would say, so I asked Claude (which I have a paid membership for and have discussed political things about many times). The answers were broadly similar: https://imgur.com/a/caFgKHH
What do you mean by America's fault? Maybe I'm in a left wing bubble, but I was under the impression that it's generally accepted by left and right that the Vietnam war was not justified and did not accomplish its stated goals.
Abortion? Permissive drug laws? Immigration? Homelessness policies? Vaccines? American AI gave me these examples, and they're not really so cut and dry. Do you have a better example that's being left out (no pun intended)?
Good Lord I hadn't ever heard of the mai lai massacre. Just spend 10 mins reading about it. Christ.
There is no “Chinese LLM”. Each “lab” is distinct and their models behavior is as unique as those from OpenAI and Anthropic
Somehow a certain set of labs are all releasing open weights and a certain other set of labs are closed weights.
Somehow the two main closed weights frontier models come from two companies with HQs about two miles apart, and the CEO of one used to work for the other.
I don't understand the premise in the beginning
how is running servers supposed to be 0 cost, while running ai inferrence isn't?
> how is running servers supposed to be 0 cost, while running ai inferrence isn't?
For a SaaS business, running servers isn't free. But compared to the cost of running GPUs for inference that you are selling, it almost is. The company I work for is a SaaS company. We have a single production server. A couple of QA servers. All hosted on Hetzner. Monthly cost for servers is less than $400. This generates a few million dollars a year in revenue.
If we were in the business of selling inference, our cost of providing the service, for the same amount of revenue would significantly higher.
Even large businesses like Microsoft, Meta, Google have operated with similar margins. Cost of running servers, compared to revenue was very low. But inference changed that, in a dramatic way.
A typical server that costs 10k to 30k to own and operate can serve between hundreds and thousands of requests per second of a traditional web application like facebook for 2-4 kW of power, the marginal cost of each request is effectively zero.
A single response from kimi k3 requires hardware that cost between 500k and 1m dollars up front and draw over 20kW. Each request costs at least 5% to 10% of the charged cost.
The real scary part is China's compute capacity is orders of magnitude smaller than US's.
Thinking machines will do it for usa
later secondaries investors in openai/anthropic. It's like time traveling into the spacex ipo.
oh wow - hadn't realized they decided to opensource Qwen 3.8 Max. That's pretty big news.
> To that end, here’s an even more interesting question around distillation: why exactly is it bad? After all, what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models that are themselves being distilled? Who is exactly being wronged here?
Frontier labs that thought they could Rupoor[0] the entire creative class, transferring the coercion premium of copyright ownership from Hollywood to themselves. In their eyes, copyright should not apply to them, but also their models should have exactly the same value as a copyrighted work.
Stratechery also argues the US should explicitly make training fair use and forbid terms of service that prohibit distillation. I'm in support of the latter, but NOT the former, even though I normally hate copyright. My reasoning is primarily that copyright is one of the few legal paths available for a rando to go and put the work of an AI frontier lab in legal jeopardy. In the EU and Japan, such legal action has already been foreclosed by similar law. And while free distillation would obviously be preferable, it's also much more of a legal long-shot. Getting America to do anything that even smells like taking property away from the powerful is impossible[1] - it's our zeroth amendment. But we can at least hack the property laws that currently exist to cause problems for the frontier labs.
And, to be clear, if distillation is OK but training is not fair use, distillation is still OK. The output of an AI model is never copyrightable, because copyright only protects the human element. Essentially, this would say "don't train on humans, but absolutely rip off and steal the shit out of other AI labs and give it to the rest of us."
[0] In the Legend of Zelda series, Rupoor is anti-money - collecting it decreases the amount of money you own. I am using it to mean "turn someone's asset into a liability".
[1] Given that America was literally created to protect a wealthy land/slave owner class from disenfranchisement, either from above or below, and the last time we did this we literally had to fight a civil war against that same owner class that installed a new owner class that has largely remained today
Honestly, as someone from a developing country, this shift is good for us. US frontier models are too expensive for us to use regularly. Chinese open-source models/subscriptions are really good to use.
A company making a decision to allow use of chinese models is a company also choosing to send tons of various credentials to chinese model companies. These will just get scooped up, OpenAI and Anthropic can probably hack into anything at this point if they wanted to.
But, they are releasing the weights very shortly (or already have for some of the models discussed). For a very large company, you can purchase or rent the hardware yourself to serve the models.
Or any US hyperscaler with GPUs to spare can decide to serve the models for a reasonable cost/token.
You don't have to send China your data.
Are the open weights models accelerating layoffs in Indian IT sector?
They are neither accelerating or decelerating layoffs in Indian IT sector, from what I know they were bound to happen as software and IT were never hard skills. They were anyway supposed to move to Africa or other countries within a decade, so a lot of executives were cautious even before 2023. AI has simply killed the opportunity for an Infosys or TCS in countries in Africa or even other South Asian countries.
The world would be so much better off if the WITCH companies ceased to exist
1 reply →
Interesting and true perspective.
Commenting wholesale on some folks who are asking for hard evidence. I cannot provide that either but can contribute some empirical data.
I have been working on a project with about a dozen generation tasks, each of which comes with a fixed token budget. The nature of this system requires that most tasks be completed by distinct model families.
As a result, I tested ~50 models across as many model families as I could gather, frontier and open weight, API (gateway and direct) and self-hosted. Evaluation was based on a set of cosine similarity validations that was repeated across ~50 different embedding models.
Interestingly, frontier models did worse on the tasks than open weight models. However, when it came to costs, the picture was reversed: frontier models were much, much more token-efficient. In fact, almost no open-weight model was able to meet the initial token budget, while almost all frontier models did. Moreover, open weight models struggled massively with reasoning, in terms of latency and token consumption.
I also found that the latest models did not perform better than older models. And any a priori benchmarking data was utterly useless.
So, I ended up using a set of open weight models without reasoning, as it turned out reasoning as well as frontier negatively correlated with the tasks. However, before I knew this, I had spent a lot of time running each available reasoning level for each model.
Lastly, as an aside, when it came to embedding models, size (dims as well as model size) did not correlate with quality, once a hurdle figure (~2k dims) was met. In fact, sweet spot was 3-5K, and for my (text-based) set of tasks, dense models tended to outperform MoE ones.
Someone (anyone!) get David Sacks on the horn and tell him to read this.
I personally believe opensource or may be state owned LLMs are future, every country on earth should have its own national LLM,trained on country's own data, and then allow its public to use it for free
I see a lot of ways how that could lead countries into a dystopian nightmare, where the gov gets to decide what kind of biases a model should have.
Gemini Pro has become so bad for my purposes -- copy & paste Go programming and code analysis with the web GUI -- that I'll take any model with equivalent capabilities at the same price or lower. I don't care where it comes from, I'm not dealing with state secrets and, frankly speaking, US corporations have an abysmal track record regarding safety and surveillance.
There are really only two factors at play here: people trying to protect their massive investments, and governments fighting over who gets backdoor access to all of your chats.
I really enjoyed reading this.
This might be a simplistic take, but my biggest worry with depending on Chinese models (and, by proxy, open-weights model development) is that the US can deem them a national security risk at basically any time, and Ant/OAI have minimal interest in making frontier-level models open-weights.
Regulated companies prohibit Chinese models in anticipation of the ban-hammer from the feds, so for data-sensitive work, they're stuck with LLaMa, gpt-oss and Gemma models (which are good and serve as a good-enough base for sft, but seemingly not as good or as expensive as Chinese models)
I suppose the USG can do the same thing that China is doing and bankroll/subsidize that effort; whether they will is for fate to decide.
Nonetheless, this article made it clear that nVIDIA is the real winner in all of this. Shovel selling to the extreme.
The fact that Anthropic has a model like Mythos means that counterpart countries like Russia and China are not far behind, if they haven't already developed something similar or better.
As a Russian: no. Sber's GigaChat 3.5 (released two weeks ago) is the best we can right now and it's years behind SOTA models.
> Russia
lmao
It's say Anthropic, Allegedly OpenAI...
All the article relies on the premise that selling tokens is profitable. I don't see any indication for this, and it makes the whole house of card crumble.
The future is SLMs and China is going there..
Is this article written by AI ? It looks so.
The american AI companies, presumably
I think the most interesting part of the article is the Huggingface incident at the end:
>Right now defenders are effectively banned from using Fable or Sol for cybersecurity because of Trump administration directives; that means the best alternative is using models from a country which has been trying to weaken our cyber defenses for years. This is insane!
I understand some guardrails are needed, but it is becoming increasing problematic manage them without a strong public discussion.
American models == Cars with no brakes. Chinese models == Cars with brakes that can't drive down tiananmen square.
>> Going forward, however, I expect the inference market to grow much faster than training costs (and that includes the assumption that training costs will continue to skyrocket), which means they really can make it up in volume.
But as inference becomes cheaper, some of the market will move to self hosted inference. I look forward to someone supplying small servers designed to run inference locally.
With or even without open models these companies are selling compute, and we've been making that rent vs buy decision for 60 years.
I am French, and I am sad to see that French models aren't being talked about and that it remains a battle between the Americans and the Chinese.
Answer: US investors
I haven't had the time to look into recent and past history, but my intuition points at failed empires having similar "elites" starting to stagnate innovation for the sake of "protectionism" - whatever that means.
Nothing changed for China. The only difference between then and now is that now China is the one selling the products, instead of western capitalists taking a cut off the COGS and selling price.
You techbros need to get off your ass and go to work.
Why hasn't someone created a disinformation benchmark to assess that part of the argument?
i think most is vcs
The options are to use LLMs from a country run by a psychopathic regime or alternatively to use Chinese LLMs.
not enough people
I'm honestly more afraid of Claude
what a horrible article. full of misinformation and dishonesty.
1. training new base models are expensive for sure, but fine-tuning them are relatively inexpensive enough the labs can continue to do so forever. the main reason why frontier models are so good is because the massive input they generated from user usage. they are using that information to strategically build better training data. and this is why no other models can catch up, til now that is. but if chinese models are good enough, and free to host, and cheaper to use, then the consequence is the frontier labs will lost valuable user inputs and the chinese labs will gain more. as time goes by this will be a domino effect.
2. nvidia is not only the player in the hardware scene. amd mi350p is getting popular, and huawei is pumping SuperPoDs. what does this mean for us? chinese models will surely use chinese hardware, and optimize for them. the other people will pick amd because compare to nvidia they are cheaper. with open weight models and open source inference stacks, they are freely to experiment and improve the stack, thus further lower the inference cost and nvidia dependency. and they even plan to build their own inference hardware, too. and nvidia loses market share meaning all the fund it gives to openai or anthropic will be cut, too.
and you say there is nothing to afraid?
> Second, intelligence isn’t in fact a perfect commodity
We have got very far from Cicero's coining of the word 'intelligentia' (from inter legere, a 'reading between' and hence discernment) when people talk about 'intelligence' as a commodity
People have been decrying the 'cheapening' of the word intelligence for over a century now, going back to Psychology's adoption of the word and coining of nonsenses like "Intelligence Quotient". "Artificial Intelligence" is just the latest degradation of the original humanistic meaning, and now people aren't ever bothering to prepend 'artificial' to their idiotic use of the word
[flagged]
[flagged]
[flagged]
[flagged]
[flagged]
[dead]
[dead]
[dead]
[flagged]
[dead]
[dead]
[dead]
[dead]
[dead]
People with 401k’s and retirees?
Looks like authored by AI. With quite some reasoning, but no real data to back up main point.
What makes the Chinese models this good? I don't believe it's distillation alone.
This from OpenAi's Head of Strategic Futures "Some observations on Kimi: It's a very good model! I don't think its performance can be explained away by distillation or anything like that"
https://x.com/deanwball/status/2078133895766114412
China's strategy of spending billions on training these models and open sourcing these models away is strategic - they want to kill the US LLM industry at any cost.
To win on the AI front by any means necessary.
> China's strategy of spending billions on training these models and open sourcing these models away is strategic - they want to kill the US LLM industry at any cost.
Why is it when Anthropic and OpenAI spend billions trying to beat each other it is competition, but when the Chinese companies do it then it is trying to kill the US LLM industry at any cost.
The US federal government spends billions in subsidies via the US Chip Act, and bans chip sales to China to support US companies.
But the implication is that somehow Chinese competition is illegitimate because "strategic".
Jingoism is the answer I believe.
I am not defending the US LLM industry or the government, all I am saying is it's similar to an arms race. I did not suggest anywhere that what China is doing is illegitimate. China has state supported capitalism, and they will do whatever it takes to prop up the AI industry. Anthropic and OpenAI want the same protections.
>What makes the Chinese models this good?
Why wouldn't it be? China is pumping out AI research and researchers at a staggering pace and there is no inherent reason why western models should be better
I think frontier labs should start building ecosystems by partnering with companies that already have software products, and even collaborating with hardware manufacturers. The ultimate goal should be to create a much broader range of products that integrate naturally into people's everyday lives.
The United States' real advantage over China is freedom. Chinese LLMs simply can't compete with American ones when it comes to the humanities, creativity, entertainment, or financial transparency. As long as the U.S. continues monetizing these strengths, the compounding effect will make it virtually impossible for China to surpass the U.S. at the product level.
This is a really ironic comment given how the US gov is getting politically involved in pretty much all science research funding and speech at the moment.
Exactly. If this continues, the US will be fighting the wrong competition against China instead of building around its own structural strengths.
> humanities, creativity, entertainment, or financial transparency
> monetizing these strengths
> real advantage over China is freedom
Please tell me if I'm unfairly paraphrasing but these seem to be your main argument and they seem to be oxymorons
[dead]
Kellogg School of Business -- he said -- token as a commodity and therefore Open AI is constrained .. ha ha ha hee hee ha .. Well... you build a better mousetrap, and DeepSeek, K3, and ByteDance are just that -- just as good and fit to purpose -- What is needed is to build on top of -- not paniteir (invade privacy and kill people with the information) -- not USMC AI -- use PI's as overwatch killer drones -- but how can I make harder steel, longer-lasting, seawater-resistant concrete, faster time to build housing, better enforcement of USDA rules and FDA adverse enforcement, and better EPA water cleanup, a better FTC for consumer goods -- that is, if I buy an item, that item is safe and built to purpose -- ANYONE not talking about public protection of consumer rights usng AI, is wasting your time
Has someone replaced your return key with a double-dash key?
ok look at this way, without being petty what is your ability to have clean water today? and how is that measured? -- You know the food is substandard to the USDA standard EG ( Taylor farms 2026 ) But Where is the enforcement for Your drinking water and the food YOU eat daily. And what is the projected outcome in years from eating substandard foods to NIH standards? -- All of this is known today. And AI can help, by YOU building on top of the models you have access to in 2026. The idea that tokens are constrained - is the wrong approach in Business and public policy. You may recall how KSG got its funding and why? In todays world YOU have a responsibility to build better - As such a better mousetrap.