Comment by OsrsNeedsf2P
13 hours ago
DS isn't better than Sol and Gemini in the benchmarks, but after working with them all for a while now I'm beginning to understand why people find DS _feels_ so much better.
First, it's the tok/s and TTFT from providers. Deepseek blazes through work - at quite good quality - and you don't find yourself waiting for your agents to finish.
Second is the openness. Being able to see the CoT and not being randomly blocked for cyber review. You're never wondering if it's stuck or if it will randomly fail. Combined with the great cost value.. My next subscription will probably be Z.ai
Feels better to me in the sense that it doesn't try to be too clever. When I ask it to do things it actually generates code that a developer would write. That still lets me build software in layers like I do mentally. And it keeps you in control of where abstractions are.
Some people would say that you don't need this at all with the frontier llms. But when extending existing software or writing software that fits into an existing ecosystem, I find I still need to offer a lot of guidance. And it's better that the context window isn't flooded with code that was reverted after I looked at it.
> Some people would say that you don't need this at all with the frontier LLMs
Hit the nail on the head. Development methodologies pivoting at the whim of frontier models have been one of the things that have been worrying me as of late.
it doesn't worry me too much tbh. This is really just cutting corners and going this route leaves you with some kind of additive ship of theseus piece of software you don't understand and require your frontier model to work with, which just... becomes more additive and gross. right now we're in some proverbial heyday of everyone seeing how absolutely grand of a thing they can make with frontier models. over time i expect some old design philosophies to take hold with people who use LLM's successfully and find a place again paired with developers who understand why such layered methodologies are important... because this whole AI thing wasn't the silver bullet CEO's and crazy execs thought it was. They will eventually get sick of the support burden of badly designed stuff. i anticipate a bit of pendulum swinging back towards "still using AI, but we need smart people to do it because it's expensive and we want it done right". Right now a lot of companies are being extremely reckless imo and they haven't fully realized/felt the cost implications of trainwreck software yet.
11 replies →
I've been using layered development techniques like the parent says even with frontier models, just one agent at a time with Zed.
I just use a suite of tests to enforce the mechanical standards (dry, encapsulation, etc) and arch documents for shape + reviews of new work against the two. I’m not seeing any significant drift.
3 replies →
> Feels better to me in the sense that it doesn't try to be too clever. When I ask it to do things it actually generates code that a developer would write. That still lets me build software in layers like I do mentally. And it keeps you in control of where abstractions are.
yes!So it's not just me. In terms of how intelligent they want to be OpenAI is at the top, then Claude. And actually when I was playing with Sol it meant that many tasks it would just try to do something else because it thought I don't know what I want. Even trying to nudge it back in the right direction wouldn't work.
I guess the models just have the cultures of their creatures.
> it doesn't try to be too clever
DS is miles better than GLM in this regard. I wanted to add that as PC is planning to go to GLM next.
Interesting. I’ve been using GLM 5.3 Flash for a couple months and found it to be great. What has it been doing that you found to be “too clever”? It sometimes gets stuck in loops. I once asked it to do a code review of a simple pattern and it went off on a 45 odyssey trying to analyze too much. But other than those, it’s been a pretty good daily workhorse. And you can’t beat the price. Even Deepseek 4.1 Flash is more expensive. And I switched to GLM 5.3 Flash from DS 4 Flash, my previous daily driver, which I also found quite good. Anyway, my point isn’t to start a “which model is better” subthread, but was just curious what behavior you’ve seen where it’s “too clever.”
5 replies →
I briefly tried GLM 5.3 Flash and the first thing I noticed was how slow it was followed quickly by its single-mindedness. With DS 4.1 Flash I find that it understands that not all the detail in my prompt is important. Sometimes I like to give a quick example of what I'm looking for and it can take the hint whereas GLM will seemingly reshape its worldview around this tiny detail.
DS just handles ambiguity and my bursts of detail with much more grace. I'm only using "max effort" and the CoT is reasonable compared to GLM.
2 replies →
What harness are you using?
> My next subscription will probably be Z.ai
Wait... you just spent 3 paragraphs singing the praises of DS... and then you say you're going to subscribe to the one that makes GLM. ?
(I do tend to agree with you, though I tend to favor GLM-5.3-flash over DS 4.1 flash.)
Z.ai does offer access to DS AFAIK, so I guess that's what they mean? Maybe the feeling is that it provides more value that DS's own platform since you also get GLM, not sure.
maybe the point was become of the openness, they will be using z.ai?
But it’s not at all as described in terms of work style - it’s much more like Claude.
1 reply →
Agreed.
It doesn’t have to be better, it just had to be nearly equivalent. I’ve been doing a lot of vulnerability work lately (trying to get ahead of the bots finding exploits in our code) and the frontier models feel like a chaperone at a church dance: get too close and they show up with the ruler and enforce the safety margin. Using deepseek and glm and qwen, I’ve gotten nearly equivalent results without the constraints. I still use frontier models for things but outside of a company environment, I’m happy to spend 1/10 the money, get the same results, and not have try and trick the AI into working.
No need to use a chinese provider, in case that's an issue for you. https://fireworks.ai provides this model for an ok price.
I'd rather my data go to China and directly to labs that release open weight models (DeepSeek is MIT licensed still); instead of closed labs and the NSA/CIA; who has far better ability to hurt me than China does.
Without getting too much into politics; over the past several decades, one country is going around starting wars, regime changes and other interventionism, or initiating economic warfare (Huawei sanctions because they were getting too competitive to US companies; GPU restrictions; tariffs; etc), and that's not one administration: it's US policy over the past few decades.
Have a look here:
https://ourworldindata.org/grapher/human-rights-index-vdem
31 replies →
The biggest problem with this is that I can't run my company's own internal benchmarks against this API, because they will be trained into the next version of the model, making them worthless.
Have you ever wondered why Manus founders gave up billions of dollars from Meta and “voluntarily” moved back to Beijing? One country does things out in the open, while the other operates in the shadows. The latter worries me more.
3 replies →
This is really a roundabout way of saying you have no plans to live, or perhaps even travel, to China.
To add to this, Fireworks (and many competitors, e.g. enumerated here[0]) offer zero-data retention inference, to anyone. DeepSeek cloud does not. It's part of their terms that anything you send them is fair game for them to train models on.
[0] https://openrouter.ai/deepseek/deepseek-v4.1-flash
(The badges next to companies indicate their privacy terms, i.e. green/yellow/red)
It’s pretty clear despite what they want you to think Anthropic and OpenAI train on everyone’s data who has a subsidized plan. Only way you get ZDR is by paying API prices. Also note that it is zero retention, not zero training.
2 replies →
I figure (hope) buying from DeepSeek directly benefits them the most so they can release even more improvements and open-weight models.
Plus only one country has threatened to annex mine and it wasn't China.
Assuming canadian or dane, but maybe there are other places he’s threatened. There are canadian inference providers running deepseek, and definitely eu ones idk about ones based in denmark.
The economic sovereignty of any nation, but especially smaller ones like canada and denmark, is harmed by sending our money out of the country when there are options available in country.
7 replies →
It surely does. Official DeepSeek provider logs and trains off your data.
China has annexed (or it claims so) several countries already, so it sounds like you're only concerned with yourself - probably not what you wanted to achieve.
(But I'd concede that in the field of genocide China doesn't come up as bad)
If one is not American, why prefer American? Chinese fine by me, I trust the US providers just as little.
There's also European providers for these models. More than one!
10 replies →
I support people making choices with full knowledge of what they are doing and what the consequences are, as much as it is possible. What I oppose is making choices without such knowledge.
There are multiple others in EU, and in US. Cheaper, more expensive, slower, faster. So much choice.
Completely unrelated, but it's so odd to see upside down '8's in the first column of the base tables on their pricing page.
The link points to OpenRouter with multiple providers.
>> Being able to see the CoT
Dont assume that is how your model decides...
https://arxiv.org/abs/2305.04388
https://arxiv.org/abs/2307.13702
https://arxiv.org/abs/2503.08679
> they frequently generate CoT explanations rationalizing those answers
I can't tell if this is a really human thing to do, or if it's just a sycophantic thing to do when your master likes a good story.
Fair, but you can see the logprobs, right? That alone is pretty valuable.
This is like one of those folk theories that have some truth but in practice it is useless.
I love COT and gives me a good idea about the thought process. Sure it doesn’t work always
Same. And the biggest benefit is that you get to see if it’s heading in the wrong detection
In practice, it does help to immediately be able to tell if you missed something or wrote something incorrect in your prompt, compared to if you didn't have access to the CoT. If you see that it goes off wondering about stuff or thinking about stuff that you know are already solved/known, you cancel the execution, edit the prompt and redo, then observe the CoT again, until you've resolved any "misdirections".
Both OpenAI and Anthropic hides so much of that, that you don't really know if it misunderstood something or you got something wrong, until all the turns are complete or you directly see tool calls that implicitly tells you something is wrong.
You don't need to read any papers to see this effect concretely when developing together with LLMs.
[dead]
Absolutely!
DeepSeek feels like an excellent sparring partner. The others feel like mind readers.
I don't have to worry about any complex context engineering / multi agent stuff if I am using DeepSeek because the sequential stuff is fast enough. I don't have to bother about touching a million context length because its cheap enough.
Where it seems to lag behind is 3d, computer use, motion graphics, design but the initial demos seem like the vision model has improved substantially.
DeepSeek v4.1 Flash hallucinates like crazy.[1]
Is DeepSeek: DeepSeek v4.1 Flash 1017 any better?
[1] - https://www.youtube.com/watch?v=P4dTq4X8bqk
>My next subscription will probably be Z.ai
please don't subject yourself to this. z.ai subscriptions are not a good value. just pay API rates, or use a subscription like opencode go.
I've used the zai subscription, opencode go und cline-pass in the last few months and would recommend the latter. Zai quickly drains your quota as soon as the context window is at more than 10-20% and opencode was a bit unreliable at times.
Is that a dig on DeepSeek not offering subscription plans?
If so OpenCode Go looks good and includes DeepSeek models (yet to try myself).
I wonder what quantization opencode uses for DS models.
From "What about data and privacy?" in FAQ it sounds like they don't host models themselves. e.g. it says:
They also follow official DeepSeek API peak/off peak hour pricing and link to their docs:
edit: they used to explicitly name the providers, at least in the past DeepSeek V4 was served with official API https://github.com/anomalyco/opencode/issues/24649#issuecomm...
Everyone wonders that.
1 reply →
Deepseek is not from Z.ai, that's GLM.
I second this, I love DS and am not looking back. For Elixir, not a lot of models are good, but DS just blazes through it.
Google's Argon model via antigravity is unusable, for a small code review I asked, already spent like 10% of the weekly usage, in less than 10 mins. Very useless imho.
I always find DS or Qwen to be very spot on on audit changes or plans for just a few cents.
I'm enjoying Alibaba's Token Plan. They don't have any hourly/daily cutoffs. You get your allotment for the month and use it when you want.
If I were part of a cabal that is interested in preserving control and trying to secure trillions of dollars, it would be a minor effort to nudge and bend the curve against competitors like the Asian AI complex.
How would you go about that?
I do not use DS for coding, but for various natural language processing tasks (even translation) in large volume, and it works wonder.
It is very capable for a lot of tasks, not only coding like the other open models.
> not being randomly blocked for cyber review
Just to test, will GLM – hosted by an infra provider other than Z.AI – also stop answering if I asked controversial questions from the pov of its country of origin? Or third party infra providers can perform some unhinging on the model before hosting? Or actually the blocking happens even before the question reaches the LLMs in that country?
(Note: my paid LLM access at this point are only via DS and GLM subs - 90-95% usage; and a free Gemini Pro access. I hope that adds some cred that I am not trying to take potshots at the their country of origin).
Depends on the question. Most models won't tell you how to make weapons because they're trained not to, but the political stuff is usually blocked on the API level
It will not. Ask DeepSeek about Tiananmen Square. It will happily answer
to add to the corrent sibling answer, the unhinging is called abliteration by the wider community but it's hard to find someone who will serve that over API.
GLM refuses to write exploits. It's not open season with anything you want.
Deepseek and Muse are my go to for cheap. both work perfect for basic and mid tier tasks
What about mimo v2.6 pro?
I’ve jumped between mimo and ds quite a bit lately. Mimo seems to be slightly better in terms of what it can achieve in terms of end results. But it gets there very slowly, and in a super convoluted way. Sometimes it’s almost frustrating to read its reasoning. Also I’ve noticed that it tends to not use tools in the harness as much as other models.
I usually default to ds, and if it gets stuck try mimo but usually just for fire and forget tasks. After a bit I come back and see what it did.
Usually great but it infuriates me sometimes with what it does. DeepSeek is more consistent and certainly faster.
I mean... I ported some of my childhood Finnish shareware games for Linux and Rust with DeepSeek V4.1 Flash. It took about six hours of spelunking and finally I was pretty happy with the results! I did it twice. Tokens for both of the games were around 3 euros.
Muse is fantastic for open source work. At least 1.3 was, for its price. Wonder makes less economic sense.
I like it because it does the thing I want it to do rather than moralizing.
No, I don't care about license provenance for all the components of a 30 line script that i'm using in my house, for myself to sync up some folders. No, i'm not going to find and archive the licenses for all the previous projects leading to this for future legal disclosure.
Yes, I own the DVD for the movie i'm transcoding. No I don't want to talk about the legality of rebroadcast.
Deepseek (and most chinese models as a whole) don't care . oAI and Anthropic models want to make it point #1.
... which is extra hilarious knowing the laws broken to create the damn models in the first place.
there's a literal meme about this floating around the harness/AI chats for the past few months[0]
[0]: https://i.ibb.co/SwH4Sgwq/796779070-28483090404651613-920066...
I don't see this kind of behavior myself, I wonder if it's me not pushing the boundaries enough or me explicitly or implicitly framing the prompt in a way that makes it less likely for the agent to intervene.
I think opencode go is the better value to get access to DeepSeek.
and really , for 90% of standard agentic work, deepseek flash is more than good enough. its the main driver for our Personal Assistant offering and users love it.
> Deepseek... My next subscription will probably be Z.ai
Wait wut? You do realize that they are two different companies, right?
Z.ai (well technically alibaba) used to support both GLM and DS models in their API and coding plans.
z.ai was not alibaba. z.ai got glm and alibaba got qwen
For us neophytes, what does this mean: “randomly blocked for cyber review”?
It happens very frequently if you happen to work on anything low level, e.g. a C code base, a database code base, etc. that deals with memory, and these have nothing to do with cyber security. It is extremely annoying and makes these models unusable.
E.g., your prompt for simple auth works (e.g., decode JWT), but your sessions stop because you are flagged for cybersecurity risk.
The Fable is the worst offender, and it is not purely Anthropic's fault.
It's happened to me exactly twice in about 6 months of using codex (via pi). Both times were completely bogus and annoying but still, it's not too bad.
1 reply →
Agreed, but don’t subscribe to z.ai; their inference is terrible. Stick with openrouter
Huh, doesn't DeepSeek take the longest because it overthinks way too much? Haven't tested this release yet but was very disappointed with past ones because it takes an eternity to do anything because it constantly overthinks and questions itself way too much instead of just starting.
It thinks a bit more, but if the cache reads are free or almost free, and it blasts through 10x faster than GPT Sol/Terra/Luna, it doesn't really matter. Just don't use the max thinking for implementation, go with high or low.
Why not use max thinking for implementation? For cost reasons?
2 replies →
I think, as long as you don't set thinking to max and use a fitting system prompt, it's not a big issue.
Yes, and this version explicitly addresses that
4.1 Flash on medium is a solid builder agent
LLMs don't think.
I love using glm 5.3 more than deepseek because of performance
Isnt Z.ai for GLM? Does Z.ai provide Deepseek?
> Being able to see the CoT and not being randomly blocked for cyber review.
I had a bad experience with MiMo from Xiaomi. I had it trigger unsafe content blocks while fixing visual bugs in a completely vanilla space game it just wrote from scratch in the same session. It also got triggered for changes in some other benign greenfield app.
also it's cheap, using it in opencode is great
Note that opencode routes deepseek requests to China.
I have to ask what's everyone's issue with the requests to China. We've had 25 years of American big tech basically capture all our data, and use that to make insane profits.
10 replies →
My code is opensource, china already has it. For my professional work we have entreprise plans with OpenAI / Anthropic
For an average European, why is it better than routing requests to the USA?
9 replies →
Opencode also has a setting to not route to China, though this would probably just disable the models that do not have providiers in alternative locations.
If you pay for a business plan on openrouter (I know, it’s not opencode, just another option), you can geo lock providers if China is a concern.
Doesn’t that depend on the router service provider?
For 99% of the tasks any model really would do, even GPT 5.4 from an year ago.
It's only for long running/very difficult cases that the differences are here, and they are small.
I can think of plenty of tasks that even the closed frontier models fail miserably at. From my perspective we have a long way to go for these models to get better.
I'm sure they exist, but it's not the bulk of the work.
2 replies →
still a myracle to me what people achieve with it? I tried so many things with it, everything is broken it produces