I seriously dont' know all this big hullabaloo about one shot prompting.
by definition, a single prompt wont' constitute the complexity of a software project. ergo, what you'll get is a series of assumptions made by the model based on preexisting code in its training corpus.
I'd rather see a coding agent that can follow steps in a plan file to a T while following guardrails and adhering to the proper coding conventions in the human reviewed spec.
Id rather see performance in agent loops against human defined objectives where it can be verified to stick to defined guardrails and continue without drift till its objectives are complete.
I'd also like to see it identify bugs and potential performance increases by identifying existing code and suggesting refactors based on context it can pickup about the particular use case you are trying to create.
These are way more valuable metrics than "hey build X"
> A policeman sees a drunk man searching for something under a streetlight and asks what the drunk has lost. He says he lost his keys and they both look under the streetlight together. After a few minutes the policeman asks if he is sure he lost them here, and the drunk replies, no, and that he lost them in the park. The policeman asks why he is searching here, and the drunk replies, "this is where the light is"
All of your suggestions are better but they're hard, so someone casually evaluating an AI isn't going to do them.
Sure, for casual evaluation, I agree. But are there serious analyses that are evaluating this kind of thing? I mean, these are the kinds of things I evaluate in my own work when a new model comes out, or when I'm evaluating a harness. But this is all very ad hoc and intuitional. I'd love to start bringing rigor to it, but I haven't found much prior art on this. In another thread someone said that's because it's probably impossible to do this rigorously because too much of it is subjective. And that does match my intuition. But I continue to suspect that intuition is wrong.
The minute an open model breaks through and beats Claude Opus/Fable, it's over.
There are far more opportunities that can be served when the world's intellectuals have the raw weights and can fine tune, splice, distill, and reapply.
Imagine having raw unfettered access to Fable. It can be refit to structural biology. It can be fine tuned on the repo for smaller context requirements. It can be run cheaper and air gapped.
It's the "starting from empty slate" greenfield that's the real problem.
We used to make fun of Engineers who follow a README on a framework, test it on an empty project, and say "this framework is the best for our 10 year running production app". Greenfield mentality is always the solution to all problems and problem to all solutions.
One should still measure oneshotting, it's an important self-measurement metric - but against an established, large codebase.
There are upcoming benchmarks aimed at measuring the ability to work with brownfield tasks. (Of course, benchmarks can be gamed, but they are still better than unrealistic toy tasks that earlier generations of benchmarks used. Frontier labs are yet to use them in their tech reports or marketing material, though.:-)
At least they did some analysis. I've couple AI slop "X is the best tool for the job" that didn't even try it. (Worse, we are already using QT which has a tool for the job, and the QT tool works with the rest of the QT ecosystem unlike whatever AI told them)
Unless I'm missing something, the prompt he gave must have been fairly detailed because both games are basically identical.
But for a more practical issue, the ultimate goal of LLMs is to replace software engineers, or at least enable everybody to become a software engineer, to use a more up-beat phrasing that's no less accurate. And so an LLM's ability to reliably construct something from a poorly defined, contradictory, or otherwise flawed prompt, while accurately inferring intent is probably the first finish line.
> I seriously dont' know all this big hullabaloo about one shot prompting.
It's a relatively objective way of testing LLMs, and I think it's pretty representative of how strong models are overall.
The outcome of this test mirrors how GLM 5.2 and Opus 4.8 work for me: they're both similarly capable of fully executing a given task, but Opus tends to have a bit more "taste" in how it handles unstated details or implicit requirements.
> what you'll get is a series of assumptions made by the model
Yes, but that's why we use these models in the first place. We don't want to explicitly write down all the details because that would mean writing code. So we write a higher-level, human-language spec, and let the LLM fill in the blanks. The question is how good they are at doing that.
I guess the experiment is interesting to determine if a model can produce something subjectively valued as "good" based on fairly vague and open-ended specifications. The benchmark is not to determine if the output fits the input, but whether the output is internally consistent: it's a game, but does it behave as one would expect that any game behaves? Does it end when you each the goal, do you die when hitting the spikes, are there weird edge cases in behavior when you move around?
I think however that they should have used the same harness and also repeated the experiment a few times to judge the variance in results.
It's a proxy for what you actually want to measure.
Note that after the model generated a bunch of (intermediary) code, they still have to have it tested and get bugs fixed (via the agent/harness). In this "one shot" you still have agent loops against human defined objectives.
And these toy examples give some insight as to how the model performs. If the test were "here's some code written by $corp, please take these tickets and work on them" it may be a "real" example but nobody would be able to make sense of actually how "hard" it is, or how "well" the model did the job, besides the workers already familiar with the context.
As someone who works at $corp - there is a massive different in tickets. I've seen "The is not spelled 'teh'", and I've seen some other service is writing to memory causing a crash in my service (the later took months to track down since our code was correct and nothing gives a hint of where to look). Both problems are important to fix, but the first is so simple I don't care how good AI is (the hard part is getting it through the process)
It's true that no one is trying to one shot anything serious right now, but it's still an important metric. Claude Code and Opus really took off when they improved the harnessing enough that it would self-correct many of its mistakes without needing user input. In fact I think long-term autonomy (in the range of several hours) and self-correcting is going to be where we see most improvements in coming years.
> In fact I think long-term autonomy (in the range of several hours) and self-correcting is going to be where we see most improvements in coming years.
Right, model intelligence defines the scope of things they can one shot
I also suspect that users naturally calibrate to a model's useful scope, gradually getting positive/negative feedback and gradually making their requests bigger/smaller than before
If a model can take a series of increasingly complex instructions and satisfy the requirements without human intervention, we can pretty easily decide how well overall the model does. And, judging better models just means adding more requirements to a task. So, I think it's a useful method (Even if it's not a realistic use case).
Of course, with a software engineer at the helm - the models are going to be able to be guided to produce much better output. (Or worse, depending on the engineer!)
You seem to be missing the point of what parent is saying :)
To really evaluate how a model is to use in real life, it should have access to tools, and be able to iterate on something, like they do when you use them in an agent harness.
None of that iteration need necessarily to have a human driving it (although if you're building something you want to be able to maintain, you probably need a human driving the design and architecture), you can just let the model do a couple of tries and give it input into how it's doing, and you get something closer to how people use these models in reality.
> If a model can take a series of increasingly complex instructions and satisfy the requirements without human intervention (...)
This is the wrong metric to target. Today's models can feel one-shot but they are so at the expense of resilient ReAct loops that brute force their way out of the mess initial prompts created.
And each iteration is expensive.
Sometimes failing fast and early is better than going for one-shot models that try to mitigate the mess they created with reasoning steps and ReAct loops.
On one hand, that's sort of true for practical uses - and benchmarks notoriously undercount multi-turn settings.
On another, being able to reliably tackle minor tasks with no handholding is very valuable in itself. Sometimes implementation details are important, but often, the most important thing is to Get It Done.
> I'd rather see a coding agent that can follow steps in a plan file to a T while following guardrails and adhering to the proper coding conventions in the human reviewed spec.
Guardrails/conventions should be enforced in linters, formatters, static analysis tooling; not specs/prompts.
lets say you have a table that is partitioned. how do you lint/format "any select into this table MUST include the partition key in the predicate and any join must include it in the on." I'm not personally familiar with any static analysis tool that does this but its trivial to implement with an llm prompt. trivially easy to add to your automated PR reviews.
It's not always possible, or at least trivial. For example how do you enforce "prefer to reuse existing code over making a copy"? Is there a static analysis tool that will detect two pieces of code that do the same thing?
When the model produces reasonable results from one prompt, you could assume that it will also return reasonable results through the follow up prompts.
The argument is flawed, there is no logical reason to assume a single prompt won’t be sufficient to constitute the complexity of a software project. It may not be practical in many cases but there is too much variability in what is considered a complex software project and in the sufficiency of instruction in a single prompt to make that claim and say it’s “by definition.”
And that prompt will basically be 2000 page spec Bible à la IBM circa 1960, see waterfall. Unless AI develops mindreading (and advanced mindreading at that), single prompt creation of actual complex software products will never happen. You'll one shot a simple non scientific calculator, but not Excel or Vim or Nginx.
One shot prompting/tooling is the only reasonable way to use an llm in my opinion. You should not be having an LLM operating for hours creating thousands of lines of new code that you can never review or maintain. You can actually be highly productive modifying a single file or two at a time, ideally as focused and little context as possible, without the llm being given full permission to add as much context as possible along the way to maximize revenue for the developers of the harness.
The agentic engineering paradigm is just a narrative trend pushed by AI companies to get people to 10x their token consumption per prompt. It plays into people's laziness and addiction to dopamine too causing addict like behavior in people that fall prey to this trend.
"We did multi-shot prompting to try and get these two games into comparable states using these two different models."
"Well obviously you provided better follow-up prompts to the one that came out better."
Also nothing about human-provided plan files and guardrails preclude the one-shot benchmark test. Heavens, I almost said "real coding," but in "real agentic program creation" you'd obviously be doing multi-turn interaction with the agent, but how can you provide a fair test when the model's output n determines your n+1 response?
Sure, real-world usage is always more difficult to benchmark, but the additional issue with the one shot prompting benchmark is that by optimizing for it, models are nudged towards making all those assumptions they shouldn't really make. Maybe a better test would be to have a fully spec'd-out plan, but start with a one shot, high-level prompt and expect the agent to discover your preferences by repeatedly asking for clarifications. The system that manages to suss out more of the details in the hidden spec this way, in less steps and with less unnecessary questions would more likely to be a truly well-calibrated agent.
Blame anthropic, they decided to make these type of one-shot examples the primary focus of the Fable 5 release, and relegating benchmark scores to the pdf.
That's precisely the difference between an engineer and a business guy.
The business guy would say "hey build me this and that" and would get _something_ to show of.
An engineer will have a long conversation with a llm about the exact requirements, tech stack, tradeoffs. He would understand what is built, how is it built, and refine on the fly until he gets something sensible.
It won't be as fast as "build this", but the result will be much better and more maintainable.
For the enginering workflow, you don't need Fable. Any model better or equivqlent to Sonnet 4.6 would do. Yes, sometimes it will hallucinate, sometimes it'll be wrong, but it's our job as engineers to correct it and have full ownership of the result.
Single prompt performance is interesting because best agentic results of yesterday turned out to be best single prompt results of today.
If we stopped developing LLMs the the only reasonable way to benchmark them would be to compare yheir performance with all the tricks we can build on top of them. Sine the are still developing rapidly any apples to apples comparison is worthwhile.
Of course this particular benchmark is not really single prompt but rather "agentic without steering".
The thing with one-shot prompting is that it tests the ability for the model to make good choices on its own, rather than only instruction following.
Instruction following has been down for years, and while there are of course metrics that continue to improve as the frontier advances (for example, the ability to continue following the original instructions even as context grows), you can't really get that much better at performing a list of instructions as-written if the instructions are sufficiently precise enough that there's no wiggle room for interpretation (which seems to be what you are describing).
For example, one of the things that got me the most excited for Fable 5 was its ability to work for over eight hours straight on a single instruction and seemingly faithfully the entire time. That was something I observed personally after trying out the same workflow that runs for maybe two or three hours with Opus and then still needs followups. Fable needed no followups. That's a game changer for me compared to the prior state of the art.
That kind of stuff is going to end up being the most beneficial to people who are touching the edges of their knowledge or even exploring completely new areas. And that type of work is exactly the kind of work that makes agentic coding so powerful, even as much as it gets harder to judge the quality of the work when you lack the skills yourself. It's a good thing that the quality increases across the board, even for skilled practitioners.
For example, even people who know how to write inference engines or how matmul kernels work or how to optimize model architecture can't always predict just the sheer breadth of things agents can try to improve performance, and sometimes you get over some wall and reach a completely different optimum that you just wouldn't have reached in any reasonable amount of time by applying traditional knowledge even if you're an expert in the field.
That kind of stuff is amazing. And that's exactly the kind of stuff that one-shot prompting is testing for. It's kind of like testing for the model's "innovation", as much of an oxymoron that is.
> So we ran it head-to-head against Claude Opus 4.8: same one-shot prompt, build a 3D platformer in raw WebGL from scratch
Running a single one-shot prompt is not a benchmark, not is it representative of any sort of real-world usage.
Most agent usage is collaborative so you need to test things like reliability (when I delegate a task, does it complete it without making up test results for e.g.) and steerability (does it obey my instructions or does it just do what it thinks is best).
Hi, I am the author, I completely agree! I set out to run a vibe test on this one, not a benchmark, the real benchmarks are listed. My test shows what the models can do when both tasked with a long-running, technically difficult, one-shot task.
I think your test you describe (collaborative, task delegation, task completion, TTD, steerability) is a great format for a future test that I will definitely try out.
Thanks, I didn't mean to be brusque, but I have seen a lot of these vibe tests lately that come to grand conclusions like "X model is better than Y" from the result of a single prompt.
Appreciate you sharing the results of your tests though!
On the other hand, I did just leave my pi agent running GPT 5.5 overnight on a clearly defined, long running task. It's been running about 10 hours now and it's mostly done. So this kind of use case is also valid.
Thinking about it, I would say that the majority of agentic work I do, by a long shot, is subagents which are launched from the main session, using a prompt of its choosing. Those could be considered short versions of these fully autonomous tasks.
Care to share more about your pi setup? I've recently started using it (after long-time Claude Code work) and was wondering how you'd achieve these long-running tasks. Do you allow it to spawn sub-agents? Thank you!
Yes, part of the reason I chose the one-shot test was really to test long-running tasks. A lot of people seem to be experimenting with this format, for example in the now trending loop-writing workflows. And really I am interested in diving into the murky waters of these novel workflows.
sure that's why we look at a mix of formal benchmarks, one longer analysis of a side-by-side, and various other people who we trust to form an opinion, all covered in the article - not intended to be a formal benchmark, there are enough of those.
I was never able to get these models to collaborate with me the way Opus does. I'm probably an outliner, I don't one-shot projects, I don't vibe code. I basically use LLMs are if I was working with a coworker, fairly smart one, but with short memory and often missing the big picture. Sometimes I can delegate more, sometimes less, but I know I always have to stay on top of what's happening, because it WILL create mess when it hits something hard. With the Antropic models, this kind of cooperation is easy (with the exception of Opus 4.6, which was bad for some reason).
To me one shot prompting is as relevant as Strava's KOM is for cycling, i'm more interested in a good cycling performance after a 3 hours ride than a straight up 30 min record effort.
I’m actually amazed at the output since GLM doesn’t have eyes. If GLM 5.2 costs 1/5 as much, seems like it could be set up to reach out to a multimodal model for vision tasks when required. Closer to parity but probably still significantly cheaper.
I've been checking out GLM 5.2 on some projects and few thoughts on it:
- it takes it sweet time to get code rolling, not the fastest model by any means
- it strays a lot during discovery/planning but then corrects
- it's not steering friendly, as it hallucinates things that it doesn't follow later on
- its output is quite good
A sample use case: I was optimizing rendering on Swift+Zig codebase. It chocked on 5k data entries.
GLM 5.2 spent 20 minutes building the benchmarks and getting data out, which made me frustrated so I blocked non-editing tool access and went AFK, after approx. 30 minutes I found that it used already-made benchmarks and some "conclusions" to optimize 3 choke points. Output pointed that it couldn't validate suspicions and asked for more data.
Implementation worked well, it was idiomatic and non-intrusive. I would even say that it was more idiomatic than GPT 5.5 effects on same repo.
I would opt in in using it more BUT GPT usually completes same requests 5x faster.
GLM 5.2 was spark for preparing and running inside isolated containers with JJ workspaces (so that multiple can be ran in parallel).
I used it the other day for something of low importance that other models simply weren't figuring out and I didn't want to burn up Opus 4.8 on. (It had to do with overriding left-click on a macOS menu bar and then making Ctrl+click or right click bring up the menu like left-click normally does, and doing all this conditionally.)
Switched the model to GLM-5.2 halfway in the middle of a troubleshooting session (didn't even bother to reprompt, just changed it in the middle of its reasoning), gave it a few minutes, problem fixed. This is with the subscription based allocation on OpenCode Go, where a problem like this would completely burn up my Opus for the current 5 hours or even the current week.
Its also nice that you can see its entire reasoning trace. I can see it going off the rails - or see something I forgot to tell it - and stop and correct it. Or I'll learn WHY it made the choice it did and not have to question it after.
Strong agree! I deeply appreciate this aspect of GLM. Watching it think & being able to nudge early is incredibly useful. Being able to point at bad assumptions is incredibly useful. Watching what it's seeing is super informative.
It's always a shock to me how opaque most other models are!
It also is pretty resilience to letting you inject in while it's working without going off course or while getting back on track after, which I appreciate
"GLM-5.2 hit a problem here, because it can't read images. It isn't multimodal. So instead of looking at a screenshot, it fell back on a hacky workaround: it wrote scripts to read the raw pixel data and check whether the colors came out roughly as expected."
Right, just give the text llm access to a vision specific agent and that problem can be solved. Or if you really want let it even call Opus with an image - seems like you’d still save money
These style of comparisons are decent at showing capability but they don't really show me what I truly want - a sounding board and implementer with senior engineer-level execution. When I look back at all the teams that I've been part of, the best outcomes came from white-boarding (sometimes in the metaphorical sense) with one or two people, at times arguing, then finally compromising on a plan. Instead of synthetic benchmarks that try to be objective, I wonder if there's a way test this, or maybe I'm opining on a way of working that will soon be gone?
I've signed up with Ollama to experiment with these open source models. For the past 3 months, it's just been experimenting, trying it out. GLM is the first model that I am using on a daily basis to do my coding work (as well as using Claude). It's good - I've been maxing out my Ollama usage limits everyday :)
So GLM emits fewer tokens and does fewer tool calls, but still takes over twice as long to complete.
Can someone explain to me where that time usage is coming from if not from the model operation itself?
Are the individual tool calls more complex and take more time to complete? Or is the rate of tok/s lower because the model does more compute per token?
I have noticed that Opus and GPT 5.5 are very good at adjusting their thinking / reasoning intensity depending on the task at hand, something the open weights models are still not as good at.
In addition to that, some of the open weights models like GLM 5.2 or DeepSeek v4 Pro tend to be MUCH slower when generating tokens, which contributes to the perceived slowness. Although I wouldn't call models like GLM 5.2 slow by any means, e.g. it is currently one of the fastest models inside Notion today.
Probably the data center where the model is running more than anything. Another option is if Opus is using anything like a Mixture of Experts approach, in which case the amount of the model loaded in memory at one time could be smaller than GLM.
I've been using GLM 5.2 extensively for the last few days. It is slower, and the lack of multimodality is a bummer.
But, it produces solid results for a fraction of the price. Worth checking out if you have the time.
One of my goto "tests" of a new frontier models is having it rebuild a programming language from scratch. For GLM 5.2 I had it rebuild the old Rebol language in Rust:
How are people running this locally? I just checked llama.cpp and it appears unsloth has a version but it hacks a bunch of things to make it work and isn't optimal.
No one is doing that for a model this size it would have to be so heavily quantized that it wouldn’t be useful - or you’d need to spend a half million dollars on hardware. People use hosted APIs. Open weight means cloud vendors can host it.
My understanding was that n-shot prompting just referred to the number of examples included in a prompt, not the number of prompts to achieve the desired result.
"Build a 3D platformer game from scratch, in raw WebGL, with no game engine or 3D library" would be a zero-shot prompt.
> GLM-5.2 cost a fraction as much. Opus finished in half the time and shipped a cleaner game.
Off topic, but does anyone else instantly pick up on LLMisms like this? It seems like all the models have converged on this style of writing, and improvements aren't really changing it.
Yep, as I reread my own sentances I notice these LLMisms and have to rewrite them quite often. Reading so much llm-output definitely impacts your writing style.
This is excellent feedback thank you! These LLMisms in writing are a challenge I am living with currently and trying to improve on. The technical writing industry is taking a huge knock right now with companies demanding more work in less time with a big drop in quality, day to day I get less and less time to work on the quality in the prose of my work. We are working at the frontier of this right now, so we are the most heavily effected, but also get to experiment with the changes first which can be both stimulating and very frustrating.
No one has really talked about hybrid and using Opus to plan and orchestrate GLMs work both through initial build and code reviews. That’s a true best of both worlds and there doesn’t need to be a winner.
This is the way but Anthropic doesn’t make it easy, so I use GPT 5.5 in that role since I can use my subscription in OpenCode or OMP.
I also use MiniMax-M3 in utility roles like explore/library tasks.
I’ve had a z.ai subscription for several months so I’m on the older pricing. I’m really not sure it would make sense to do this at current rates - I could bump my Codex plan instead.
I mostly use Opus for skill development. Once I have a solid skill implementation with a good eval, I move ongoing execution to a cheaper model running under Goose. With the eval you can see if the cheaper model works well enough.
GLM 5.2 has one big issue that will limit its meaningful success and that's the value of their coding subscription.
Yes, in terms of API pricing, GLM 5.2 outperforms the competition. But the only people that use API billing for their coding work are large corporations, where these highly subsidized subscriptions are being fazed out.
At the same time, none of these companies will use a Chinese API for their employees.
For individuals and smaller teams, Z.ai's coding subscription is outperformed by Anthropic and OpenAI. You probably get around the same usage with Claude, but Codex definitely offers more usage for the amount you pay.
We can have a debate how much Z.ai closed the gap to GPT5.5 and Opus 4.8, but if I can freely decide between them in a world where they all cost the same, I simply wouldn't choose GLM.
So the important question becomes: How good will the offering from Z.ai get with GLM 5.3 or 6 and how much will OpenAI and Anthropic cripple their current offering in the near future.
Taking a view from outside the USA, European companies just had Fable taken away due to US export controls, and before that Anthropic announced it is holding their data for 30 days. There is immediate value to these firms to build their infrastructure around an AI that won’t be pulled away from them. And outside of Europe, other countries are more price sensitive and don’t have the same fear of building relationships with Chinese companies.
There is no such thing as a relationship with "chinese companies". In China there is just the State, and that is it.
If the world needs any more evidence of Europe's short-sightedness, it would be them running to China to spite the US (instead of creating fertile grounds for their own tech).
> My impression is that individual subscriptions are the loss leading hook
Except there is no evidence of this at all, just people comparing API and subscription pricing. The leaked financial info for OpenAI shows inference is profitable right now, though it does not show a distinction between subscription and API revenue... but if subscription revenue was so lossy, it would hard for total inference to still be profitable.
To be clear, I agree with this and they have my unlimited support pushing for relevance of open source models. GLM 5.2 is amazing and I couldn't be more excited.
I just think that as of today, most people will not find a good reason to switch to GLM.
This is an important point. I suspect API pricing will eventually disappear just like how paying for an MMS disappeared. It's an antiquated model. The bulk of the work is being done on "coding plans" is my wild guess.
It's annoying that the plans are so restrictive beyond usage limits. Understandable maybe, but annoying. In practice, only Anthropic (and maybe Google) are really restrictive though. They really scared me away with their policy of charging API rates after the fact if they consider your usage not TOS-aligned. This might be an ungrounded fear that I have, but I feel this is something they'd do so they scared me away.
> But the only people that use API billing for their coding work are large corporations
As well as people using 3rd party harnesses like OpenCode.
> At the same time, none of these companies will use a Chinese API for their employees
So who are Amazon Bedrock (who serve GLM) targetting?
Individuals are presumably going with one of the cheaper US providers such as DeepInfra ($0.18/M cached input for GLM vs $0.50 for Opus) or Fireworks AI.
> At the same time, none of these companies will use a Chinese API for their employees.
nice try but you intentionally ignored the entire Chinese market & Chinese big corporates. there are 130 Chinese companies in the fortune 500 list, with an average revenue of 80 billion USD each. do you think they are going to sign up for Claude, Codex or GLM? now consider South East Asia, Africa, Middle East, Middle Asia and South America, tell me why their large corporates won't be using GLM API billings?
your western centric view of the world is totally out of date, like it or not, 2026 is vastly different from 1996, the US no longer controls high tech whatsoever.
Also, I was testing out the GLM 5.2 using Openrouter because that's where I've got an account with some money and then when I wanted to perhaps subscribe for a better deal at z.ai, their infra was clearly overloaded to the point the 5.2 was timing out on 100% of chat requests, so perhaps I will try later when the infrastructure catches up with the model capability. Only then I can make sure their subscription is worth it.
I'm on glm pro subscription and I get so so so much more usage than Claude or Codex! I hammer on glm all day. It's a more expensive plan, but I would need a much much much bigger plan for codex or Claude to do what I do.
> Through an API it costs a fraction of Opus, and you can run it yourself for free if you have the hardware.
I haven't been keeping up on hardware costs for state of the art LLM inference, but this remark made me ask myself how many readers of the article would actually be able to run this model on hardware they own. How much would it cost to acquire such a setup?
This framing local LLMs as free is stupid. Basically pay 100+ months worth of API costs up front isn't free in the slightest. And it will be slower than non-local, your hardware will be outdated in 12 months and probably won't be able to run SOTA at anywhere near non-local speed in max 20 months
Yeah, it glosses over a gigantic capital expenditure. It's sort of like saying that an open source modern CPU architecture allows you to build your own CPU "for free" (provided that you own and operate a fab).
GLM-5.2 performing like it would from a good provider - 8x B200s, so $450k. (No personal experience here)
GLM-5.2, severely quantised, 512GB Mac Studio, somewhere between $10k-$35k for a used M3. Or run it on a CPU with 768GB of RAM by getting an old PowerEdge with DDR4 for around $5,000.
Qwen-3.6-35b-q6, runs well on an RTX 5090 ($4000 + cost of a PC), runs medicore on an Intel Arc B70 ($1000 + cost of a PC plus lots of fiddling to get the setup to work right).
Gemma is a good candidate for the cheaper stuff, but I lack personal experience with using it locally
>On output tokens, GLM-5.2 is less than a fifth the price of Opus.
Opus is most expensive model in pay as you go model, but IMO fair comparison should include subscription price as well. For example when one has $100 Claude Max and use it up through the month, it might not be more expensive than GLM, or at least not 5x.
I’m currently trying to figure out whether a downgrade from Max 5x to Pro in combination with one of those would save me money and if so, how much.
Edit: seems like Anthropic Pro + GLM Pro (Yearly) would let me almost halve my costs of Anthropic Max 5x. Only concerns are about GLM 5.2 not having vision support and also being kinda slower and also not being as good as Opus.
I'm considering shifting to the OpenAI $20 plan + GLM. OAI has the best computer use, vision support and the best programming intelligence of any model short of Mythos/Fable, and the quota is a lot more generous than the Anthropic $20 plan.
Yes this is true. This test was run on a $20 pro Claude subscription. I would definitely love to try use both models on the highest plans for a whole month and compare the two, great format for a future head-to-head comparison.
Cost difference matters most as cost optimization is the whole point of AI. Time difference (30 min vs 1 hr) is not a deal-breaker. The small precision gap on the first iteration does not matter for 99% of the work that happens in real world.
I was surprised today by how much better GLM-5.2 was than GPT-5.5 at aesthetic/UI work. I'll keep my Claude/Codex setup via Conductor for now, but this model got me to set up OpenCode, download their desktop app and do most of my work there today.
You should repeat this experiment but with progressively more detail in the initial prompt. Claude's secret sauce is taking weakly specified prompts and making passable things from them, but as the degrees of freedom in the prompt go down Claude starts to disobey while other models close in on the intent.
Nice comparison, but perhaps a more informative one would be to keep the harness the same and use Claude Code for both model. In your comparison, the differences could be due to many harness design decisions.
I may be biased and interested as I'm going to give you an affiliate link, but really honestly Synthetic LLM provider is a beast! They provide perfect GLM5.2, awesome token/s, TTFT and price.
2016 me would agree, but 2026 me looks at Trump and Dario, and at China, sees basically no ethical difference (or possibly even an ethical deficit for America) and considers that perhaps it's better to go with the option that isn't trying to hoodwink me with bullshit platitudes and flag waiving while doing whatever they want in actuality.
GLM-5.2 is quietly becoming the most interesting open model release this year. The coding benchmarks are surprisingly close to frontier models at a fraction of the inference cost.
We've had the great small Qwen 3.6 early April that many could actually run on their laptop. Then similar from Google a few weeks later (Gemma4, better in prose, worse in code). Then the super cheap large Deepseek V4 a few weeks later. Then antirez DS4 build that made that actually runnable on MacBooks and Mac Studios. And now the "near-frontier / near-Opus" GLM 5.2.
For people who follow open LLMs, none of these were quiet and all were the most interesting open model release for a few days/weeks. In one or two months, it will be some other model again. Now I do appreciate the real rapid improvements in open models. But there's also a ton of hype and fast-fashion around all of this.
The difference here is that those small models are impressive, but not super useful. Deepseek 4 is impressively cheap for the intelligence, but not reliable enough to daily drive unless your time has low value.
GLM passes a meaningful threshold of reliability/utility that puts it in a different category for real work. Just like Opus really took off after passing a threshold with 4.5. It's the first open model to do that.
To me DS 4 is still the most interesting due to much lower costs. Also DS 4 training isn't done yet.
From my Opus vs DS 4 Pro personal benchmarks, 16 different real-life work tasks, DS 4 has performed as well as Opus 4.8 high overall but with few drawbacks:
- on the 16 tasks, one needed several prompts to be steered back into the topic
- its review capabilities seem much worse
- DS4 had the cleanly better solution in 3 cases out of 16, with Opus "only" doing cleanly better 2 times out of 16. But still, I want to emphasize, is the worst case scenarios that imho matter the most, not the best ones, and on that front Opus outperformed.
That being said I spent less than 2$ of API working 4 days, which is more or less what I would've spent with Anthropic APIs for less than one task.
I used GLM 5.0/5.1/5.2 for some projects, and for me, the area in which they lag behind frontier models the most are user interfaces. They get really close to Opus when it comes to pure algorithms, but when I need something like web application or a mobile app that looks and works well, they are very noticeably worse than even Sonnet.
I've seen glm 5.2 struggle writing simple compilable c code. It might be good at web, but it's world knowledge is limited due to the small model size, making it's use quite limited in my opinion.
glm-5.2 is very good if you have a good harness and workflow to use it with. in fact, i'd call it good enough if you are a software engineer who knows what you want. it writes the code. i'm wondering if i need anthropic's models at all at this point, or openai. and surely in a year we won't need them at all. Opus 4.5+ was the turning point for me, and now these open models are just as good. i don't get how you IPO these companies when their only winning product is coding agents and the competition is just as good for 1/4 the price.
What would the best way to use these open source models for a price similar to what I could pay for the cheapest plan with claude and openai ?
I would like to give them a try but I certainly not have the money to get a system able to run them, and I don't really want to pay more than the state of the art
Totally agree witg the general assessment.
The biggest problem with Z.ai model for a long time is not quality, but the inference speed and general capacity availability.
Hopefully with this recent hype, there will be more provider on openrouter for 5.2.
If you are a real engineer and uses the LLM as a pair programmer instead of delegating everything to it, even GLM 4.7 was already good enough to help you with a lot of work.
I used it with Cerebras inference at a time when it had a good coding plan at a low price, and delivered tons of stuff using it.
Having issues with coding a render for good looking realistic smoke coming off burning incense, opus 4.8 & gpt-5.5 both have code issues, glm-5.2 did it. Amazing.
The real time 3d fluid dynamics appear to be the tricky part, I wish I still had opus access, would love to see if it can do it.
I wonder how much tokens and time where used for the verifying part.
Maybe GLM 5.2 instantly found the "solution" to read the screen pixel by pixel, but it could also have been a major token and time consumer.
Hi, author here, I cannot give an exact number for how many token the verification step took, but the verification GLM 5.2 ran was very stupid and definitely a waste of time. It read the pixel color data to try and verify the scene rendered properly. Which is really bad. Opus opened the game in a Playwright browser and took screenshots to verify the actual image. Which helped a lot.
Pro tip: You could use a multi-modal model to verify images as a subagent spawned by GLM 5.2, to get around this issue.
Still on a z.ai legacy plan and their 50% discount for switching to standard plans tips the balance for me. So I guess I’ll reevaluate round about beginning 2028…
I'm really feeling a bit tired of these models. I feel that since opus 4.1, I haven't been able to clearly feel the intelligence improvement from the model upgrades (except for gpt 5.5 and opus4.6 being able to speak like a human)
Seeing the results I don't see how the results are even comparable Opus is clearly far superior in most aspects. Smoothness, design, functionality etc.
At the end of the day, the time earned is more important then the cost for big players.
The ability to spawn 10 claude agents and rush a project to outcompete someone is more important for big businesses in my imo. Also the small details that GLM missed would take significant more time to iron out, considering it already took double the time.
I do hope other (open weight) models catch up, but to act like they are anywhere close for me is a bit disingenuous.
Not my experience at all, I have written about comparing DS4 vs Opus 4.8 on 16 real life work tasks on multiple posts.
Also, every single lab does RL on benchmarks, which is why Opus 4.6 was the last truly great assistant, after it, all models tend to drift into implementation asap.
I seriously dont' know all this big hullabaloo about one shot prompting.
by definition, a single prompt wont' constitute the complexity of a software project. ergo, what you'll get is a series of assumptions made by the model based on preexisting code in its training corpus.
I'd rather see a coding agent that can follow steps in a plan file to a T while following guardrails and adhering to the proper coding conventions in the human reviewed spec.
Id rather see performance in agent loops against human defined objectives where it can be verified to stick to defined guardrails and continue without drift till its objectives are complete.
I'd also like to see it identify bugs and potential performance increases by identifying existing code and suggesting refactors based on context it can pickup about the particular use case you are trying to create.
These are way more valuable metrics than "hey build X"
The streetlight effect:
> A policeman sees a drunk man searching for something under a streetlight and asks what the drunk has lost. He says he lost his keys and they both look under the streetlight together. After a few minutes the policeman asks if he is sure he lost them here, and the drunk replies, no, and that he lost them in the park. The policeman asks why he is searching here, and the drunk replies, "this is where the light is"
All of your suggestions are better but they're hard, so someone casually evaluating an AI isn't going to do them.
Sure, for casual evaluation, I agree. But are there serious analyses that are evaluating this kind of thing? I mean, these are the kinds of things I evaluate in my own work when a new model comes out, or when I'm evaluating a harness. But this is all very ad hoc and intuitional. I'd love to start bringing rigor to it, but I haven't found much prior art on this. In another thread someone said that's because it's probably impossible to do this rigorously because too much of it is subjective. And that does match my intuition. But I continue to suspect that intuition is wrong.
2 replies →
The minute an open model breaks through and beats Claude Opus/Fable, it's over.
There are far more opportunities that can be served when the world's intellectuals have the raw weights and can fine tune, splice, distill, and reapply.
Imagine having raw unfettered access to Fable. It can be refit to structural biology. It can be fine tuned on the repo for smaller context requirements. It can be run cheaper and air gapped.
The world wants this.
2 replies →
[dead]
Feels a rather outdated little parable, since nowadays one would expect the police officer to either arrest or shoot the person.
2 replies →
IMHO, It's not the oneshotting.
It's the "starting from empty slate" greenfield that's the real problem.
We used to make fun of Engineers who follow a README on a framework, test it on an empty project, and say "this framework is the best for our 10 year running production app". Greenfield mentality is always the solution to all problems and problem to all solutions.
One should still measure oneshotting, it's an important self-measurement metric - but against an established, large codebase.
There are upcoming benchmarks aimed at measuring the ability to work with brownfield tasks. (Of course, benchmarks can be gamed, but they are still better than unrealistic toy tasks that earlier generations of benchmarks used. Frontier labs are yet to use them in their tech reports or marketing material, though.:-)
* SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarios https://arxiv.org/abs/2512.18470 * SWE-CI: Evaluating Agent Capabilities in Maintaining Codebases via Continuous Integration https://arxiv.org/abs/2603.03823
At least they did some analysis. I've couple AI slop "X is the best tool for the job" that didn't even try it. (Worse, we are already using QT which has a tool for the job, and the QT tool works with the rest of the QT ecosystem unlike whatever AI told them)
Unless I'm missing something, the prompt he gave must have been fairly detailed because both games are basically identical.
But for a more practical issue, the ultimate goal of LLMs is to replace software engineers, or at least enable everybody to become a software engineer, to use a more up-beat phrasing that's no less accurate. And so an LLM's ability to reliably construct something from a poorly defined, contradictory, or otherwise flawed prompt, while accurately inferring intent is probably the first finish line.
> I seriously dont' know all this big hullabaloo about one shot prompting.
It's a relatively objective way of testing LLMs, and I think it's pretty representative of how strong models are overall.
The outcome of this test mirrors how GLM 5.2 and Opus 4.8 work for me: they're both similarly capable of fully executing a given task, but Opus tends to have a bit more "taste" in how it handles unstated details or implicit requirements.
> what you'll get is a series of assumptions made by the model
Yes, but that's why we use these models in the first place. We don't want to explicitly write down all the details because that would mean writing code. So we write a higher-level, human-language spec, and let the LLM fill in the blanks. The question is how good they are at doing that.
I guess the experiment is interesting to determine if a model can produce something subjectively valued as "good" based on fairly vague and open-ended specifications. The benchmark is not to determine if the output fits the input, but whether the output is internally consistent: it's a game, but does it behave as one would expect that any game behaves? Does it end when you each the goal, do you die when hitting the spikes, are there weird edge cases in behavior when you move around?
I think however that they should have used the same harness and also repeated the experiment a few times to judge the variance in results.
It's a proxy for what you actually want to measure.
Note that after the model generated a bunch of (intermediary) code, they still have to have it tested and get bugs fixed (via the agent/harness). In this "one shot" you still have agent loops against human defined objectives.
And these toy examples give some insight as to how the model performs. If the test were "here's some code written by $corp, please take these tickets and work on them" it may be a "real" example but nobody would be able to make sense of actually how "hard" it is, or how "well" the model did the job, besides the workers already familiar with the context.
At least everyone knows what a 3D game is.
As someone who works at $corp - there is a massive different in tickets. I've seen "The is not spelled 'teh'", and I've seen some other service is writing to memory causing a crash in my service (the later took months to track down since our code was correct and nothing gives a hint of where to look). Both problems are important to fix, but the first is so simple I don't care how good AI is (the hard part is getting it through the process)
It's true that no one is trying to one shot anything serious right now, but it's still an important metric. Claude Code and Opus really took off when they improved the harnessing enough that it would self-correct many of its mistakes without needing user input. In fact I think long-term autonomy (in the range of several hours) and self-correcting is going to be where we see most improvements in coming years.
> In fact I think long-term autonomy (in the range of several hours) and self-correcting is going to be where we see most improvements in coming years.
Right, model intelligence defines the scope of things they can one shot
I also suspect that users naturally calibrate to a model's useful scope, gradually getting positive/negative feedback and gradually making their requests bigger/smaller than before
it wont happen, its all a money grab.
1 reply →
One shotting is useful to test but only with a huge prompt (eg, build something according to this spec).
I agree generating millions of tokens from a handful of input tokens doesn't convey anything meaningful to me.
> a single prompt wont' constitute the complexity of a software project.
The top agent is for steering, but all subagents are mostly oneshot prompts
If a model can take a series of increasingly complex instructions and satisfy the requirements without human intervention, we can pretty easily decide how well overall the model does. And, judging better models just means adding more requirements to a task. So, I think it's a useful method (Even if it's not a realistic use case).
Of course, with a software engineer at the helm - the models are going to be able to be guided to produce much better output. (Or worse, depending on the engineer!)
You seem to be missing the point of what parent is saying :)
To really evaluate how a model is to use in real life, it should have access to tools, and be able to iterate on something, like they do when you use them in an agent harness.
None of that iteration need necessarily to have a human driving it (although if you're building something you want to be able to maintain, you probably need a human driving the design and architecture), you can just let the model do a couple of tries and give it input into how it's doing, and you get something closer to how people use these models in reality.
> If a model can take a series of increasingly complex instructions and satisfy the requirements without human intervention (...)
This is the wrong metric to target. Today's models can feel one-shot but they are so at the expense of resilient ReAct loops that brute force their way out of the mess initial prompts created.
And each iteration is expensive.
Sometimes failing fast and early is better than going for one-shot models that try to mitigate the mess they created with reasoning steps and ReAct loops.
I think you're underestimating the elegance of "hey build X". It already captures a lot of what you're interested in.
Additionally, with "Hey build X" nobody is happy with the methodology and people rightfully complain about the set up.
Using your suggestion the methodology would require a lot of presumptions & arguments regarding why you choose it and think it relevant to people.
Either people would not "get" it quickly enough or would disagree/not be interested on the setup because its not how they use LLMs.
On one hand, that's sort of true for practical uses - and benchmarks notoriously undercount multi-turn settings.
On another, being able to reliably tackle minor tasks with no handholding is very valuable in itself. Sometimes implementation details are important, but often, the most important thing is to Get It Done.
> I'd rather see a coding agent that can follow steps in a plan file to a T while following guardrails and adhering to the proper coding conventions in the human reviewed spec.
Guardrails/conventions should be enforced in linters, formatters, static analysis tooling; not specs/prompts.
lets say you have a table that is partitioned. how do you lint/format "any select into this table MUST include the partition key in the predicate and any join must include it in the on." I'm not personally familiar with any static analysis tool that does this but its trivial to implement with an llm prompt. trivially easy to add to your automated PR reviews.
1 reply →
Wrong, custom "specs" i.e. schemas, are literally all we have for "real" guardrails with LLMs.
https://developers.openai.com/api/docs/guides/structured-out...
Nothing else operates on the logprobs level and literally bans continuations that fail your schema.
2 replies →
It's not always possible, or at least trivial. For example how do you enforce "prefer to reuse existing code over making a copy"? Is there a static analysis tool that will detect two pieces of code that do the same thing?
1 reply →
I also love the term zero-shot in the AI benchmark world. So logical. So intuitive.........
When the model produces reasonable results from one prompt, you could assume that it will also return reasonable results through the follow up prompts.
The argument is flawed, there is no logical reason to assume a single prompt won’t be sufficient to constitute the complexity of a software project. It may not be practical in many cases but there is too much variability in what is considered a complex software project and in the sufficiency of instruction in a single prompt to make that claim and say it’s “by definition.”
And that prompt will basically be 2000 page spec Bible à la IBM circa 1960, see waterfall. Unless AI develops mindreading (and advanced mindreading at that), single prompt creation of actual complex software products will never happen. You'll one shot a simple non scientific calculator, but not Excel or Vim or Nginx.
3 replies →
One shot prompting/tooling is the only reasonable way to use an llm in my opinion. You should not be having an LLM operating for hours creating thousands of lines of new code that you can never review or maintain. You can actually be highly productive modifying a single file or two at a time, ideally as focused and little context as possible, without the llm being given full permission to add as much context as possible along the way to maximize revenue for the developers of the harness.
The agentic engineering paradigm is just a narrative trend pushed by AI companies to get people to 10x their token consumption per prompt. It plays into people's laziness and addiction to dopamine too causing addict like behavior in people that fall prey to this trend.
9 replies →
"We did multi-shot prompting to try and get these two games into comparable states using these two different models."
"Well obviously you provided better follow-up prompts to the one that came out better."
Also nothing about human-provided plan files and guardrails preclude the one-shot benchmark test. Heavens, I almost said "real coding," but in "real agentic program creation" you'd obviously be doing multi-turn interaction with the agent, but how can you provide a fair test when the model's output n determines your n+1 response?
Sure, real-world usage is always more difficult to benchmark, but the additional issue with the one shot prompting benchmark is that by optimizing for it, models are nudged towards making all those assumptions they shouldn't really make. Maybe a better test would be to have a fully spec'd-out plan, but start with a one shot, high-level prompt and expect the agent to discover your preferences by repeatedly asking for clarifications. The system that manages to suss out more of the details in the hidden spec this way, in less steps and with less unnecessary questions would more likely to be a truly well-calibrated agent.
Blame anthropic, they decided to make these type of one-shot examples the primary focus of the Fable 5 release, and relegating benchmark scores to the pdf.
That's precisely the difference between an engineer and a business guy.
The business guy would say "hey build me this and that" and would get _something_ to show of.
An engineer will have a long conversation with a llm about the exact requirements, tech stack, tradeoffs. He would understand what is built, how is it built, and refine on the fly until he gets something sensible.
It won't be as fast as "build this", but the result will be much better and more maintainable.
For the enginering workflow, you don't need Fable. Any model better or equivqlent to Sonnet 4.6 would do. Yes, sometimes it will hallucinate, sometimes it'll be wrong, but it's our job as engineers to correct it and have full ownership of the result.
what you said above is only true when the AI is not as smart/professional/knowledgable as that engineer.
1 reply →
Single prompt performance is interesting because best agentic results of yesterday turned out to be best single prompt results of today.
If we stopped developing LLMs the the only reasonable way to benchmark them would be to compare yheir performance with all the tricks we can build on top of them. Sine the are still developing rapidly any apples to apples comparison is worthwhile.
Of course this particular benchmark is not really single prompt but rather "agentic without steering".
Yet this is how virtually everybody is benchmarking and fine tuning.
Since Opus 4.6 I've seen later Anthropic models being more and more capable on one hand, but also less useful on multi turn open tasks.
It feels like with each model they are more and more prone to go "their own way" and jump into the implementation as soon as they can.
I can't but blame it on benchmarks and fine tuning around prompt-to-solution work.
I think that’s the point of the Superpowers SKILL
The thing with one-shot prompting is that it tests the ability for the model to make good choices on its own, rather than only instruction following.
Instruction following has been down for years, and while there are of course metrics that continue to improve as the frontier advances (for example, the ability to continue following the original instructions even as context grows), you can't really get that much better at performing a list of instructions as-written if the instructions are sufficiently precise enough that there's no wiggle room for interpretation (which seems to be what you are describing).
For example, one of the things that got me the most excited for Fable 5 was its ability to work for over eight hours straight on a single instruction and seemingly faithfully the entire time. That was something I observed personally after trying out the same workflow that runs for maybe two or three hours with Opus and then still needs followups. Fable needed no followups. That's a game changer for me compared to the prior state of the art.
That kind of stuff is going to end up being the most beneficial to people who are touching the edges of their knowledge or even exploring completely new areas. And that type of work is exactly the kind of work that makes agentic coding so powerful, even as much as it gets harder to judge the quality of the work when you lack the skills yourself. It's a good thing that the quality increases across the board, even for skilled practitioners.
For example, even people who know how to write inference engines or how matmul kernels work or how to optimize model architecture can't always predict just the sheer breadth of things agents can try to improve performance, and sometimes you get over some wall and reach a completely different optimum that you just wouldn't have reached in any reasonable amount of time by applying traditional knowledge even if you're an expert in the field.
That kind of stuff is amazing. And that's exactly the kind of stuff that one-shot prompting is testing for. It's kind of like testing for the model's "innovation", as much of an oxymoron that is.
> So we ran it head-to-head against Claude Opus 4.8: same one-shot prompt, build a 3D platformer in raw WebGL from scratch
Running a single one-shot prompt is not a benchmark, not is it representative of any sort of real-world usage.
Most agent usage is collaborative so you need to test things like reliability (when I delegate a task, does it complete it without making up test results for e.g.) and steerability (does it obey my instructions or does it just do what it thinks is best).
Hi, I am the author, I completely agree! I set out to run a vibe test on this one, not a benchmark, the real benchmarks are listed. My test shows what the models can do when both tasked with a long-running, technically difficult, one-shot task.
I think your test you describe (collaborative, task delegation, task completion, TTD, steerability) is a great format for a future test that I will definitely try out.
Tbf, most of the "real benchmarks" have issues that are just as bad. Assessing LLM performance is just hard
1 reply →
Thanks, I didn't mean to be brusque, but I have seen a lot of these vibe tests lately that come to grand conclusions like "X model is better than Y" from the result of a single prompt.
Appreciate you sharing the results of your tests though!
1 reply →
On the other hand, I did just leave my pi agent running GPT 5.5 overnight on a clearly defined, long running task. It's been running about 10 hours now and it's mostly done. So this kind of use case is also valid.
Thinking about it, I would say that the majority of agentic work I do, by a long shot, is subagents which are launched from the main session, using a prompt of its choosing. Those could be considered short versions of these fully autonomous tasks.
Care to share more about your pi setup? I've recently started using it (after long-time Claude Code work) and was wondering how you'd achieve these long-running tasks. Do you allow it to spawn sub-agents? Thank you!
1 reply →
Yes, part of the reason I chose the one-shot test was really to test long-running tasks. A lot of people seem to be experimenting with this format, for example in the now trending loop-writing workflows. And really I am interested in diving into the murky waters of these novel workflows.
sure that's why we look at a mix of formal benchmarks, one longer analysis of a side-by-side, and various other people who we trust to form an opinion, all covered in the article - not intended to be a formal benchmark, there are enough of those.
Then maybe you should add that caveat emptor to the article?
You make a very strong claim at the end that the hype is mostly real, and making it clear to what extent your claim holds should help the reader.
1 reply →
Totally agree, a single one-shot prompt can't prove anything.
I was never able to get these models to collaborate with me the way Opus does. I'm probably an outliner, I don't one-shot projects, I don't vibe code. I basically use LLMs are if I was working with a coworker, fairly smart one, but with short memory and often missing the big picture. Sometimes I can delegate more, sometimes less, but I know I always have to stay on top of what's happening, because it WILL create mess when it hits something hard. With the Antropic models, this kind of cooperation is easy (with the exception of Opus 4.6, which was bad for some reason).
To me one shot prompting is as relevant as Strava's KOM is for cycling, i'm more interested in a good cycling performance after a 3 hours ride than a straight up 30 min record effort.
I’m actually amazed at the output since GLM doesn’t have eyes. If GLM 5.2 costs 1/5 as much, seems like it could be set up to reach out to a multimodal model for vision tasks when required. Closer to parity but probably still significantly cheaper.
I've been checking out GLM 5.2 on some projects and few thoughts on it:
- it takes it sweet time to get code rolling, not the fastest model by any means
- it strays a lot during discovery/planning but then corrects
- it's not steering friendly, as it hallucinates things that it doesn't follow later on
- its output is quite good
A sample use case: I was optimizing rendering on Swift+Zig codebase. It chocked on 5k data entries.
GLM 5.2 spent 20 minutes building the benchmarks and getting data out, which made me frustrated so I blocked non-editing tool access and went AFK, after approx. 30 minutes I found that it used already-made benchmarks and some "conclusions" to optimize 3 choke points. Output pointed that it couldn't validate suspicions and asked for more data.
Implementation worked well, it was idiomatic and non-intrusive. I would even say that it was more idiomatic than GPT 5.5 effects on same repo.
I would opt in in using it more BUT GPT usually completes same requests 5x faster.
GLM 5.2 was spark for preparing and running inside isolated containers with JJ workspaces (so that multiple can be ran in parallel).
I used it the other day for something of low importance that other models simply weren't figuring out and I didn't want to burn up Opus 4.8 on. (It had to do with overriding left-click on a macOS menu bar and then making Ctrl+click or right click bring up the menu like left-click normally does, and doing all this conditionally.)
Switched the model to GLM-5.2 halfway in the middle of a troubleshooting session (didn't even bother to reprompt, just changed it in the middle of its reasoning), gave it a few minutes, problem fixed. This is with the subscription based allocation on OpenCode Go, where a problem like this would completely burn up my Opus for the current 5 hours or even the current week.
Its also nice that you can see its entire reasoning trace. I can see it going off the rails - or see something I forgot to tell it - and stop and correct it. Or I'll learn WHY it made the choice it did and not have to question it after.
Strong agree! I deeply appreciate this aspect of GLM. Watching it think & being able to nudge early is incredibly useful. Being able to point at bad assumptions is incredibly useful. Watching what it's seeing is super informative.
It's always a shock to me how opaque most other models are!
It also is pretty resilience to letting you inject in while it's working without going off course or while getting back on track after, which I appreciate
1 reply →
This mirrors my experience. I have been using it in Pi. It is smart and output is good but it is not efficient in getting there.
which thinking level? max or high?
Also pricing, I wanted to give a try, but when pricing is only 30% cheaper than Opus, I wouldn't go for it with these issues.
It's pricing is a lot cheaper if you can run it yourself.
What?
It is less than 20% of the cost of Opus at API rates. 1.40/4.40 vs 5/25.
2 replies →
"GLM-5.2 hit a problem here, because it can't read images. It isn't multimodal. So instead of looking at a screenshot, it fell back on a hacky workaround: it wrote scripts to read the raw pixel data and check whether the colors came out roughly as expected."
A better way would be to use https://github.com/openbmb/MiniCPM-V
Right, just give the text llm access to a vision specific agent and that problem can be solved. Or if you really want let it even call Opus with an image - seems like you’d still save money
These style of comparisons are decent at showing capability but they don't really show me what I truly want - a sounding board and implementer with senior engineer-level execution. When I look back at all the teams that I've been part of, the best outcomes came from white-boarding (sometimes in the metaphorical sense) with one or two people, at times arguing, then finally compromising on a plan. Instead of synthetic benchmarks that try to be objective, I wonder if there's a way test this, or maybe I'm opining on a way of working that will soon be gone?
I've signed up with Ollama to experiment with these open source models. For the past 3 months, it's just been experimenting, trying it out. GLM is the first model that I am using on a daily basis to do my coding work (as well as using Claude). It's good - I've been maxing out my Ollama usage limits everyday :)
Cool to hear, what kind of tasks have you been using GLM for? And what other models have you found useful through Ollama?
So the benchmark is : Two models with different harness produced very different results .
Glm game was completely broken Opus game was at first glance ok but also with bugs
Different models with different cost produced different non perfect results . How is it “close” ? :)
Also on costs : glm burns more tokens on average vs opus . Gpt5.5 burns less surprisingly
So GLM emits fewer tokens and does fewer tool calls, but still takes over twice as long to complete.
Can someone explain to me where that time usage is coming from if not from the model operation itself?
Are the individual tool calls more complex and take more time to complete? Or is the rate of tok/s lower because the model does more compute per token?
I have noticed that Opus and GPT 5.5 are very good at adjusting their thinking / reasoning intensity depending on the task at hand, something the open weights models are still not as good at.
In addition to that, some of the open weights models like GLM 5.2 or DeepSeek v4 Pro tend to be MUCH slower when generating tokens, which contributes to the perceived slowness. Although I wouldn't call models like GLM 5.2 slow by any means, e.g. it is currently one of the fastest models inside Notion today.
Probably the data center where the model is running more than anything. Another option is if Opus is using anything like a Mixture of Experts approach, in which case the amount of the model loaded in memory at one time could be smaller than GLM.
Could just be infra. I'm betting Anthropic is much better prepared.
I've been using GLM 5.2 extensively for the last few days. It is slower, and the lack of multimodality is a bummer.
But, it produces solid results for a fraction of the price. Worth checking out if you have the time.
One of my goto "tests" of a new frontier models is having it rebuild a programming language from scratch. For GLM 5.2 I had it rebuild the old Rebol language in Rust:
https://github.com/mhs/rebol-clone-glm-5.2
It did a fairly good job roughing in the language for a low token cost.
How are people running this locally? I just checked llama.cpp and it appears unsloth has a version but it hacks a bunch of things to make it work and isn't optimal.
https://github.com/ggml-org/llama.cpp/issues/24730
No one is doing that for a model this size it would have to be so heavily quantized that it wouldn’t be useful - or you’d need to spend a half million dollars on hardware. People use hosted APIs. Open weight means cloud vendors can host it.
Can you recommend any US based cloud providers?
1 reply →
My understanding was that n-shot prompting just referred to the number of examples included in a prompt, not the number of prompts to achieve the desired result.
"Build a 3D platformer game from scratch, in raw WebGL, with no game engine or 3D library" would be a zero-shot prompt.
> GLM-5.2 cost a fraction as much. Opus finished in half the time and shipped a cleaner game.
Off topic, but does anyone else instantly pick up on LLMisms like this? It seems like all the models have converged on this style of writing, and improvements aren't really changing it.
I think a bunch of real humans started to adopt the LLMs writing style.
Yep, as I reread my own sentances I notice these LLMisms and have to rewrite them quite often. Reading so much llm-output definitely impacts your writing style.
Indeed. I'm trying to develop a similar style. The phrasing in the quoted passage is really tight.
This is excellent feedback thank you! These LLMisms in writing are a challenge I am living with currently and trying to improve on. The technical writing industry is taking a huge knock right now with companies demanding more work in less time with a big drop in quality, day to day I get less and less time to work on the quality in the prose of my work. We are working at the frontier of this right now, so we are the most heavily effected, but also get to experiment with the changes first which can be both stimulating and very frustrating.
Yes, and it's really grating. It's like half of all new writing is done in the same "voice" now.
No one has really talked about hybrid and using Opus to plan and orchestrate GLMs work both through initial build and code reviews. That’s a true best of both worlds and there doesn’t need to be a winner.
This is the way but Anthropic doesn’t make it easy, so I use GPT 5.5 in that role since I can use my subscription in OpenCode or OMP.
I also use MiniMax-M3 in utility roles like explore/library tasks.
I’ve had a z.ai subscription for several months so I’m on the older pricing. I’m really not sure it would make sense to do this at current rates - I could bump my Codex plan instead.
I mostly use Opus for skill development. Once I have a solid skill implementation with a good eval, I move ongoing execution to a cheaper model running under Goose. With the eval you can see if the cheaper model works well enough.
GLM 5.2 has one big issue that will limit its meaningful success and that's the value of their coding subscription.
Yes, in terms of API pricing, GLM 5.2 outperforms the competition. But the only people that use API billing for their coding work are large corporations, where these highly subsidized subscriptions are being fazed out.
At the same time, none of these companies will use a Chinese API for their employees.
For individuals and smaller teams, Z.ai's coding subscription is outperformed by Anthropic and OpenAI. You probably get around the same usage with Claude, but Codex definitely offers more usage for the amount you pay.
We can have a debate how much Z.ai closed the gap to GPT5.5 and Opus 4.8, but if I can freely decide between them in a world where they all cost the same, I simply wouldn't choose GLM.
So the important question becomes: How good will the offering from Z.ai get with GLM 5.3 or 6 and how much will OpenAI and Anthropic cripple their current offering in the near future.
Taking a view from outside the USA, European companies just had Fable taken away due to US export controls, and before that Anthropic announced it is holding their data for 30 days. There is immediate value to these firms to build their infrastructure around an AI that won’t be pulled away from them. And outside of Europe, other countries are more price sensitive and don’t have the same fear of building relationships with Chinese companies.
There is no such thing as a relationship with "chinese companies". In China there is just the State, and that is it.
If the world needs any more evidence of Europe's short-sightedness, it would be them running to China to spite the US (instead of creating fertile grounds for their own tech).
1 reply →
And you have that guarantee from Xi?
My impression is that individual subscriptions are the loss leading hook. The money is made on Enterprise token contracts.
Employees and students used to coding with thousands of dollars worth of tokens (on a 20/100 dollar plan) will push enterprise to spend.
Having a Chinese model that is competitive won't displace this enterprise spend. But an open model hosted in the US/EU might.
The existence of GLM 5.2 puts a ceiling on how much OpenAI/Anthropic can charge for API Access.
> My impression is that individual subscriptions are the loss leading hook
Except there is no evidence of this at all, just people comparing API and subscription pricing. The leaked financial info for OpenAI shows inference is profitable right now, though it does not show a distinction between subscription and API revenue... but if subscription revenue was so lossy, it would hard for total inference to still be profitable.
4 replies →
> The existence of GLM 5.2 puts a ceiling on how much OpenAI/Anthropic can charge for API Access.
I believe this is the reason why we can even have this debate. Without this kind of competition we would not have these subsidies.
To be clear, I agree with this and they have my unlimited support pushing for relevance of open source models. GLM 5.2 is amazing and I couldn't be more excited.
I just think that as of today, most people will not find a good reason to switch to GLM.
The value of these models is that you can run them on your own hardware.
A company can buy a NVIDIA B300 and serve it's developers in house with unlimited tokens.
This is an important point. I suspect API pricing will eventually disappear just like how paying for an MMS disappeared. It's an antiquated model. The bulk of the work is being done on "coding plans" is my wild guess.
It's annoying that the plans are so restrictive beyond usage limits. Understandable maybe, but annoying. In practice, only Anthropic (and maybe Google) are really restrictive though. They really scared me away with their policy of charging API rates after the fact if they consider your usage not TOS-aligned. This might be an ungrounded fear that I have, but I feel this is something they'd do so they scared me away.
> But the only people that use API billing for their coding work are large corporations
As well as people using 3rd party harnesses like OpenCode.
> At the same time, none of these companies will use a Chinese API for their employees
So who are Amazon Bedrock (who serve GLM) targetting?
Individuals are presumably going with one of the cheaper US providers such as DeepInfra ($0.18/M cached input for GLM vs $0.50 for Opus) or Fireworks AI.
> At the same time, none of these companies will use a Chinese API for their employees.
nice try but you intentionally ignored the entire Chinese market & Chinese big corporates. there are 130 Chinese companies in the fortune 500 list, with an average revenue of 80 billion USD each. do you think they are going to sign up for Claude, Codex or GLM? now consider South East Asia, Africa, Middle East, Middle Asia and South America, tell me why their large corporates won't be using GLM API billings?
your western centric view of the world is totally out of date, like it or not, 2026 is vastly different from 1996, the US no longer controls high tech whatsoever.
Also, I was testing out the GLM 5.2 using Openrouter because that's where I've got an account with some money and then when I wanted to perhaps subscribe for a better deal at z.ai, their infra was clearly overloaded to the point the 5.2 was timing out on 100% of chat requests, so perhaps I will try later when the infrastructure catches up with the model capability. Only then I can make sure their subscription is worth it.
I'm on glm pro subscription and I get so so so much more usage than Claude or Codex! I hammer on glm all day. It's a more expensive plan, but I would need a much much much bigger plan for codex or Claude to do what I do.
> Through an API it costs a fraction of Opus, and you can run it yourself for free if you have the hardware.
I haven't been keeping up on hardware costs for state of the art LLM inference, but this remark made me ask myself how many readers of the article would actually be able to run this model on hardware they own. How much would it cost to acquire such a setup?
This framing local LLMs as free is stupid. Basically pay 100+ months worth of API costs up front isn't free in the slightest. And it will be slower than non-local, your hardware will be outdated in 12 months and probably won't be able to run SOTA at anywhere near non-local speed in max 20 months
Yeah, it glosses over a gigantic capital expenditure. It's sort of like saying that an open source modern CPU architecture allows you to build your own CPU "for free" (provided that you own and operate a fab).
True. But there are other meanings of "free". I.e. nobody can say "from now on you no longer have access to model X because you're an asshole"
2 replies →
GLM-5.2 performing like it would from a good provider - 8x B200s, so $450k. (No personal experience here)
GLM-5.2, severely quantised, 512GB Mac Studio, somewhere between $10k-$35k for a used M3. Or run it on a CPU with 768GB of RAM by getting an old PowerEdge with DDR4 for around $5,000.
Qwen-3.6-35b-q6, runs well on an RTX 5090 ($4000 + cost of a PC), runs medicore on an Intel Arc B70 ($1000 + cost of a PC plus lots of fiddling to get the setup to work right).
Gemma is a good candidate for the cheaper stuff, but I lack personal experience with using it locally
The price of a small house.
Practically nobody.
>On output tokens, GLM-5.2 is less than a fifth the price of Opus.
Opus is most expensive model in pay as you go model, but IMO fair comparison should include subscription price as well. For example when one has $100 Claude Max and use it up through the month, it might not be more expensive than GLM, or at least not 5x.
There is, for example, OpenCode Go subscription, which for $10 a month gives you a decently generous quota of GLM-5.2, among other models.
And z.ai themselves also have subscriptions.
to be exact, it gives you USD 60 of usage of open models.
> For example when one has $100 Claude Max and use it up through the month, it might not be more expensive than GLM, or at least not 5x.
https://z.ai/subscribe
I’m currently trying to figure out whether a downgrade from Max 5x to Pro in combination with one of those would save me money and if so, how much.
Edit: seems like Anthropic Pro + GLM Pro (Yearly) would let me almost halve my costs of Anthropic Max 5x. Only concerns are about GLM 5.2 not having vision support and also being kinda slower and also not being as good as Opus.
I'm considering shifting to the OpenAI $20 plan + GLM. OAI has the best computer use, vision support and the best programming intelligence of any model short of Mythos/Fable, and the quota is a lot more generous than the Anthropic $20 plan.
Yes this is true. This test was run on a $20 pro Claude subscription. I would definitely love to try use both models on the highest plans for a whole month and compare the two, great format for a future head-to-head comparison.
Is it fair when the one is heavily subsidized and the other one is not?
I think it's most fair to compare the plain token pricing that is used by everyone.
> Is it fair when the one is heavily subsidized
As a consumer, yes, it's totally fair. All that matters to me is the price I pay at the pump, not whether that price is "real" or not.
Z.ai is also believed to be "subsidised". Its parent company is running at a large loss right now.
Anthropic have claimed they expect their first profitable quarter this year -- they may have bigger margins on their raw API than you realise.
3 replies →
GLM has subscription plans too.
Out of stock, unavailable
Cost difference matters most as cost optimization is the whole point of AI. Time difference (30 min vs 1 hr) is not a deal-breaker. The small precision gap on the first iteration does not matter for 99% of the work that happens in real world.
Yes I 100% agree. Time-taken can be improved (with harnesses, subagent workflows etc.) and varies based on task.
I was surprised today by how much better GLM-5.2 was than GPT-5.5 at aesthetic/UI work. I'll keep my Claude/Codex setup via Conductor for now, but this model got me to set up OpenCode, download their desktop app and do most of my work there today.
You should repeat this experiment but with progressively more detail in the initial prompt. Claude's secret sauce is taking weakly specified prompts and making passable things from them, but as the degrees of freedom in the prompt go down Claude starts to disobey while other models close in on the intent.
That is a great suggestion that I am definitely going to look into, thanks!
Nice comparison, but perhaps a more informative one would be to keep the harness the same and use Claude Code for both model. In your comparison, the differences could be due to many harness design decisions.
I signed up for GLM 5.2 yesterday to try it out because Anthropic kept throwing 529 Overloaded
I like it, but the lite plan ate 22% usage of my 5h reset window in a single session after 2 prompts on xhigh of GLM 5.2 [1m]
Result was satisfactory, I think stuff is decent, I'm happy to use either, wish there was a combined subscription plan where I could get both
I may be biased and interested as I'm going to give you an affiliate link, but really honestly Synthetic LLM provider is a beast! They provide perfect GLM5.2, awesome token/s, TTFT and price.
Coupled with a local Headroom (https://github.com/headroomlabs-ai/headroom) you'll be able to use a LOT without hitting your 5h window :)
Definitely the best $ value for me considering the reasonable performance of GLM5.2.
They provide a rolling window quota, so you're never really out of quota contrary to other providers, you can adjust day to day.
Check it out if interested : https://synthetic.new/?referral=kwjqga9QYoUgpZV
---
Docs & all models : https://dev.synthetic.new/docs/api/models
Pretty clearly it's beating Opus at [web dev](https://www.gptbased.com/) - on price, on score.. I mean what else is there?
Article states it's not multimodal. I guess that means for webdev it means you can't take a screenshot to indicate errors etc.
I hate to be that guy, but real privacy policy on training data/it being hosted somewhere where I'm not worried about secrets being stored/leaked.
Open weights win on that front surely?
3 replies →
2016 me would agree, but 2026 me looks at Trump and Dario, and at China, sees basically no ethical difference (or possibly even an ethical deficit for America) and considers that perhaps it's better to go with the option that isn't trying to hoodwink me with bullshit platitudes and flag waiving while doing whatever they want in actuality.
Realistically you’d need to rotate secrets anyway once it moves from dev to production regardless of model provider
Its on other providers, like Together.ai
Latency? Just saying there's other things to consider.
> 256 GiB unified RAM.
So, 8000$, plus it's unavailable. 3 years of Codex/Opus subscription.
> API prices
Which are irrelevant for 200$ Codex/Opus plans that are times cheaper.
GLM-5.2 is quietly becoming the most interesting open model release this year. The coding benchmarks are surprisingly close to frontier models at a fraction of the inference cost.
We've had the great small Qwen 3.6 early April that many could actually run on their laptop. Then similar from Google a few weeks later (Gemma4, better in prose, worse in code). Then the super cheap large Deepseek V4 a few weeks later. Then antirez DS4 build that made that actually runnable on MacBooks and Mac Studios. And now the "near-frontier / near-Opus" GLM 5.2.
For people who follow open LLMs, none of these were quiet and all were the most interesting open model release for a few days/weeks. In one or two months, it will be some other model again. Now I do appreciate the real rapid improvements in open models. But there's also a ton of hype and fast-fashion around all of this.
The difference here is that those small models are impressive, but not super useful. Deepseek 4 is impressively cheap for the intelligence, but not reliable enough to daily drive unless your time has low value.
GLM passes a meaningful threshold of reliability/utility that puts it in a different category for real work. Just like Opus really took off after passing a threshold with 4.5. It's the first open model to do that.
1 reply →
To me DS 4 is still the most interesting due to much lower costs. Also DS 4 training isn't done yet.
From my Opus vs DS 4 Pro personal benchmarks, 16 different real-life work tasks, DS 4 has performed as well as Opus 4.8 high overall but with few drawbacks:
- on the 16 tasks, one needed several prompts to be steered back into the topic
- its review capabilities seem much worse
- DS4 had the cleanly better solution in 3 cases out of 16, with Opus "only" doing cleanly better 2 times out of 16. But still, I want to emphasize, is the worst case scenarios that imho matter the most, not the best ones, and on that front Opus outperformed.
That being said I spent less than 2$ of API working 4 days, which is more or less what I would've spent with Anthropic APIs for less than one task.
Are these games supposed to be a good example of quality output? If this is the product, I don't really want to play _either_ of them.
I used GLM 5.0/5.1/5.2 for some projects, and for me, the area in which they lag behind frontier models the most are user interfaces. They get really close to Opus when it comes to pure algorithms, but when I need something like web application or a mobile app that looks and works well, they are very noticeably worse than even Sonnet.
not apples to apples. comparing official vs. pi.dev+openrouter and having slow times is more a openrouter issue. try comparing using official z.ai.
The model is 756B parameters, open weights.
I've seen glm 5.2 struggle writing simple compilable c code. It might be good at web, but it's world knowledge is limited due to the small model size, making it's use quite limited in my opinion.
glm-5.2 is very good if you have a good harness and workflow to use it with. in fact, i'd call it good enough if you are a software engineer who knows what you want. it writes the code. i'm wondering if i need anthropic's models at all at this point, or openai. and surely in a year we won't need them at all. Opus 4.5+ was the turning point for me, and now these open models are just as good. i don't get how you IPO these companies when their only winning product is coding agents and the competition is just as good for 1/4 the price.
What would the best way to use these open source models for a price similar to what I could pay for the cheapest plan with claude and openai ?
I would like to give them a try but I certainly not have the money to get a system able to run them, and I don't really want to pay more than the state of the art
Great article,
My only, I guess feedback, is that it's not really clear about the price.
Would the 21.92 be the API pricing I guess?
Cost $5.39 (real billed) ~$21.92 (estimate, list pricing)
The text only part is the catch for me.
If it builds a UI and can't look at it, it's askin ls whether the app looks right.
Would have run it with GLM on max/xhigh effort. Just for fun.
Totally agree witg the general assessment. The biggest problem with Z.ai model for a long time is not quality, but the inference speed and general capacity availability. Hopefully with this recent hype, there will be more provider on openrouter for 5.2.
If you are a real engineer and uses the LLM as a pair programmer instead of delegating everything to it, even GLM 4.7 was already good enough to help you with a lot of work.
I used it with Cerebras inference at a time when it had a good coding plan at a low price, and delivered tons of stuff using it.
Having issues with coding a render for good looking realistic smoke coming off burning incense, opus 4.8 & gpt-5.5 both have code issues, glm-5.2 did it. Amazing.
The real time 3d fluid dynamics appear to be the tricky part, I wish I still had opus access, would love to see if it can do it.
You mean Fable?
I wonder how much tokens and time where used for the verifying part. Maybe GLM 5.2 instantly found the "solution" to read the screen pixel by pixel, but it could also have been a major token and time consumer.
Hi, author here, I cannot give an exact number for how many token the verification step took, but the verification GLM 5.2 ran was very stupid and definitely a waste of time. It read the pixel color data to try and verify the scene rendered properly. Which is really bad. Opus opened the game in a Playwright browser and took screenshots to verify the actual image. Which helped a lot.
Pro tip: You could use a multi-modal model to verify images as a subagent spawned by GLM 5.2, to get around this issue.
I could be wrong but I believe this is a non-vision model. Please weigh in to correct me bc I would love to be wrong
GLM 5.2 is text only, not multi modal. And Opus is multi modal.
Still on a z.ai legacy plan and their 50% discount for switching to standard plans tips the balance for me. So I guess I’ll reevaluate round about beginning 2028…
In the name of science we crafted an autonomous AI agent that builds games on a loop. It is based on GLM 5.2.
I am not sure where this is going to lead us but it is fun to watch.
For those praising GLM 5.2, can anyone confirm?
Tried with 2 harnesses and it seems bad + slow
Just that their Coding Plan is too hard to get. I've been trying to grab it for a week and still can't get it
While this is interesting, one single sample with different coding harness is not very scientific.
Yes I agree 100%. My next guide would do better to use identical harnesses.
I'm really feeling a bit tired of these models. I feel that since opus 4.1, I haven't been able to clearly feel the intelligence improvement from the model upgrades (except for gpt 5.5 and opus4.6 being able to speak like a human)
i think GLM 5.2 is not cheap and not easy to get the coding plan... so even it's on the Opus level... still not attractive.
I used GLM (4.7, 5 and 5.1) both through their coding plan and now through OpenCode zen. Both where painless to get.
How is it not easy to get the coding plan?
opencode zen go?
GLM is the most overrated LLM. I tried it and it not good.
GLM cannot use vision like Opus can. This is not a useful comparison.
I see your point. Just the fact that one model does have vision and one does not might be an interesting point of comparison, however.
What is this fashion of testing models by giving them one shot projects? Especially games. this is so stupid
Seeing the results I don't see how the results are even comparable Opus is clearly far superior in most aspects. Smoothness, design, functionality etc.
At the end of the day, the time earned is more important then the cost for big players.
The ability to spawn 10 claude agents and rush a project to outcompete someone is more important for big businesses in my imo. Also the small details that GLM missed would take significant more time to iron out, considering it already took double the time.
I do hope other (open weight) models catch up, but to act like they are anywhere close for me is a bit disingenuous.
I swear, if I read forms with “genuinely” one more time I am gonna scream. FUCK LLM WRITING
I hear you
[flagged]
[dead]
[dead]
Chinese models optimize for benchmarks and do poorly in real-world tasks
Not my experience at all, I have written about comparing DS4 vs Opus 4.8 on 16 real life work tasks on multiple posts.
Also, every single lab does RL on benchmarks, which is why Opus 4.6 was the last truly great assistant, after it, all models tend to drift into implementation asap.
Hi, author here, can you link? I would love to read about this.
1 reply →