I'm not sure what people are on in the comments. It doesn't beat the other models, but it sure competes despite its size.
GLM 5.1 is an excellent model, but even at Q4 you're looking at ~400GB.
Kimi K2.5 is really good too, and at Q4 quantization you're looking at almost ~600GB.
This model? You can run it at Q4 with 70GB of VRAM. This is approaching consumer level territory (you can get a Mac Studio with 128GB of RAM for ~3500 USD).
For the Claude-pilled people, I don't know if you only run Opus but when I was on the Pro plan Sonnet was already extremely capable. This beats the latest Sonnet while running locally, without anyone charging you extra for having HERMES.md in your repo, or locking you out of your account on a whim.
Mistral has never been competitive at the frontier, but maybe that is not what we need from them. Having Pareto models that get you 80% of the frontier at 20% of the cost/size sounds really good to me.
That is insane, if you billed me an extra $200 for a bug in your system I'd flat out cancel my subscription. If you're not going to credit that back to me, you don't deserve anymore of my money. I'm a Claude first guy, but if you're going to bill me incorrectly, that's on you, own it, fix it.
Where? All I see is Boris saying "we are unable to issue compensation for degraded service or technical errors that result in incorrect billing routing".
Keep this in mind next time you hear someone talking about "removing the human in the loop".
Anthropic apparently won't take responsibility for issues their own systems handling billing cause. You think they'll take responsibility in your system when a bug in their models can be demonstrated as the cause?
> Anthropic apparently won't take responsibility for issues their own systems handling billing cause.
I think with every org, especially the big ones, trying to dodge responsibility (setting the intent of "customer support" to be annoying them enough for them to buzz off), the only recourse people have is to give them enough bad press where they wake up and do the refund, it's less than a rounding error for them.
I think Anthropic is hardly unique in that position and being able to chat with a human with any sort of power to actually make things right is becoming more and more rare. If any human eyes saw that, the correct thing to do would probably be passing the message up the chain like "Hey, this will have really bad optics if we don't do the right thing. Can you take like 5 minutes and hit the refund button while I draft up a nice message about it?"
> This model? You can run it at Q4 with 70GB of VRAM. This is approaching consumer level territory (you can get a Mac Studio with 128GB of RAM for ~3500 USD).
The one thing I would want everyone curious about local LLMs to know is that being able to run a model and being able to run a model fast are two very different thresholds. You can get these models to run on a 128GB Mac, but we need to first tell if Q4 retains enough quality (models have different sensitivities to quantization) and how fast it runs.
For running async work and background tasks the prompt processing and token generation speeds matter less, but a lot of Mac Studio buyers have discovered the hard way that it's not going to be as responsive as working with a model hosted in the cloud on proper hardware.
For most people without hard requirements for on-site processing, the best use case for this model would be going through one of the OpenRouter hosted providers for it and paying by token.
> This beats the latest Sonnet while running locally
Almost every open weight model launch this year has come with claims that it matches or exceeds Sonnet. I've been trying a lot of them and I have yet to see it in practice, even when the benchmarks show a clear lead.
> The one thing I would want everyone curious about local LLMs to know is that being able to run a model and being able to run a model fast are two very different thresholds. You can get these models to run on a 128GB Mac, but we need to first tell if Q4 retains enough quality (models have different sensitivities to quantization) and how fast it runs.
Very valid. This is an active area of research, and there are a lot of options to try out already today.
- People have successfully used TurboQuant to quantize model weights (TQ3_4S), not just the context KV, to achieve smaller sizes than Q4 (~3.5 bpw) with much better PPL and faster decoding.
- Importance-weighted quantization (e.g. IQ4) also provides way better PPL, KDL, etc. at the same size as a Q4.
- DFlash (block diffusion for speculative decoding) needs a good drafting model compatible with the big model, but can provide an uplift up to 5x in decoding (although usually in the 2-2.5x range)
- Forcing a model's thinking to obey a simple grammar has been shown to improve results with drastically lower thinking output (faster effective result generation) although that has been more impactful on smaller models.
We should be skeptical, but it's definitely trending in the right direction and I wouldn't be surprised if we are indeed able to run it at acceptable speeds.
> Almost every open weight model launch this year has come with claims that it matches or exceeds Sonnet. I've been trying a lot of them and I have yet to see it in practice, even when the benchmarks show a clear lead.
This hasn't been my experience. After Anthropic's started their shenanigans I've switched to exclusively using open-weights models via OpenRouter and OpenCode and I can't really tell a difference (for better or for worse).
Cloud hardware is not inherently more "proper" than what's being proposed here, there's nothing wrong per se about targeting slower inference speeds in an on prem single-user context.
> Cloud hardware can run the original model. Quantization will reduce quality.
New models are often being released in quantized format to begin with. This is true of both Kimi and the new DeepSeek V4 series. There is no "original model", the model is generated using Quantization Aware Training (QAT).
> There is no "original model", the model is generated using Quantization Aware Training (QAT).
The original model is the model used for the benchmarks
People will say "You can run it locally!" then show the benchmarks of the original model, but what they really mean is that you can run a heavily quantized adaptation of the model which has difference performance characteristics.
That remark was specific to newer models like Kimi 2.x and DeepSeek V4 series, and this is clearly stated in my comment.
As for other models, we quantize them because we are generally constrained by the model's total footprint in bytes, and running a larger model that's been quantized to fit in the same footprint as a smaller one improves performance compared to a smaller original, generally up to Q4 or so, with even tighter quantizations (up to Q2) being usable for some uses such as general chat.
The quantization for some models can be very detrimental and their quality can drop considerably from the posted benchmarks which are probably at bf16, this is why having considerable RAM can be important.
> For the Claude-pilled people, I don't know if you only run Opus but when I was on the Pro plan Sonnet was already extremely capable.
Before February I was able to use Opus on High exclusively on my Max plan no problem. Now I've shifted to just using Sonnet on high and yeah, its pretty capable. I love that, Claude Pilled. ;)
Yeah I love Claude, amazing models. Anthropic has very quickly burned most of the goodwill I had for it so I still ended up cancelling my subscription.
Yeah, you can run it locally if you have enough VRAM, but the reports trickling in are saying about 3 tok/sec. This was on a Strix Halo box which definitely has the needed VRAM, but isn't going to have as high mem bandwidth as a GPU card, it's going to be similar on a Mac - that's the dilemma... the unified memory machines have the VRAM, but the bandwidth isn't great for running dense models. This size of a dense model is only going to be runnable (usefully) by very few people who have multiple GPU cards with enough memory to add up to about 70GB.
I don't think this is quite correct, a Strix Halo box usually has 256 GB/s memory bandwidth. An M5 Max has 614 GB/s. An M3 Ultra (no M4 or M5 Ultra) has 820 GB/s. It's still not GDDR or HBM territory, but still significantly faster.
That's the edge of Apple Silicon for AI. When they scale up the chip they add more memory controllers which adds more channels and more bandwidth.
But yeah in the end it's still going to be only a handful of people that can run it.
What I meant is that I think researching and developing smaller more powerful model is more interesting than chasing the next 3T parameter model while burning through VC money and squeezing your customer base more and more aggressively.
That's more a testament of how good Qwen3.6 27B is (it really is great) more than how bad this one is IMO. Gemma 4 31B was already good, but Qwen3.6 27B is incredible for its size.
DeepSeek v4 Flash is still over 100GB at Q4 IIRC, and Q4 has generally been the sweet spot. Although it's an MoE so it might run a lot faster that this dense Mistral model if you have the RAM.
You could run it on a single Mac Studio with M3 Ultra, or two Mac Studios with M4 Max at higher perf than that. And lightly quantizing this could give us modern dense models in the ~80GB size range, which is a very compelling target.
Wouldn't matter much still. M3 ultra has 819GB/s unified memory bandwidth. That means theoretical max tokem rate is 819/128 =~ 6.39 t/s. At 80 GB (5 bit quantization), its still near about 10 t/s ... far from a good coding experience. Also, these are theoretical max.. real world token generation rates would be at least 15-20% less.
Very valid. Importance-weighted quantization and TurboQuant on model weights can reduce loss a lot compared to "traditional" Q4 so one can be hopeful.
> For $3500 I can get 7-8 years of GLM using coding plans, have a faster model and much better code quality
But you will own no computer, and that's also assuming prices stay what they are. Anyway my point was not whether or not it makes financial sense for everyone. A lot of people are very happy not owning their movies, software, games, cars or house. I'm just happy there is a future where the people can own and locally run the tech that was trained on their stolen data.
I recommend using OpenRouter (openrouter.ai). Basically a broker between inference providers and you which allows you to pick, try, and switch models from a massive catalog, extremely transparent about usage and pricing.
I would love to be able to run frontier locally, but I think the larger importance of open weight models is price accountability.
In the US with our broken system of capitalism, it’s the only way we can tether these companies to reality. Left to their own devices, I’m not convinced they would actually compete with each other on price.
Buy nobody like to talk about how “moat” building is fundamentally anti-competitive, even in name.
Funny that self proclaimed capitalists hate the system in practice. Commodity pricing is what truly terrifies them.
I'm not necessarily interested in having frontier locally. You don't need to be frontier to be a very good and useful coding agent. I agree with your point on price accountability though. Hopefully no tariff comes down on the Chinese and European open-weight models.
As always, rooting for these guys — model and national diversity is great. This looks like a solid foundation to build on; hopefully the 3.6/3.7 will dial in more gains. It looks like maybe from the computer use benchmarks that their vision pipeline could use improvement, but that’s just speculation.
The different results on some benchmarks vibes as if this is truly an independently trained model, not just exfiltrated frontier logs, which I think is also really important - having different weight architectures inside a particular model seems like a benefit on its own when viewed from a global systems architecture perspective.
For it's size, that's really good! Though I bet it being a dense model probably helps a lot, if it was MoE at that size, I bet the benchmark performance would go quite a bit down (which consequently would also mean that I'd at least be able to run it with decent tokens/second, with the bunch of Nvidia L4 cards available to me, which presently are only okay with MoE models).
It's cool that they added comparisons to their own Mistral Small 4 119B A7B, which kind of shows that! They could have also included comparisons to something like Qwen Coder Next 80B A3B (or maybe the newer Qwen 3.6 35B A3B, or the 27B dense one), maybe DeepSeek V4 Flash 284B A13B, or the older GPT-OSS 120B A5B to illustrate that difference and where their model sits even better, it would probably give a more positive picture than just comparing themselves against a bunch of bigger models!
Come to think of it, alongside throwing some money at DeepSeek not just Anthropic, I probably should get a Mistral subscription as well sometime, to see how they perform on various tasks - cause they seem pretty cost effective and it's nice to support at least some EU orgs: https://mistral.ai/pricing
Mistral continuing to ship credible models is good for the market. Buyers need more than a two-company choice if they want pricing and deployment leverage.
Compared to all other hosted LLMs that I have tested, Mistral seems to be the only one with rather strict CSP headers. When you ask them to create a website with some javascript library it will not preview, even though le chat offers canvas mode.
Sometimes when a new release comes around from any provider I just want to test it a bit on the web. without paying and using an agent harness.
I know SVGs may not be the best benchmark, but that matches my experience of trying to run a (previous) Mistral model in Mistral Vibe, asking it to help me configure an MCP server in Vibe. It confidently explained that MCP is the MineCraft Protocol and then began a search of my computer looking for Minecraft binaries.
I can't speak for what you consider sucking, but there is a significant difference between Mistral and Kimi or Gemini. I find the others to be usable for my needs.
I agree there is a difference but does that translate to anything? It's not the same operations used to write code, and it's kind of useless. I wouldn't waste my power bill ensuring a model I was releasing was good at it.
Where are the competitive models from Singapore, Japan, Taiwan, Korea, Russia, Canada, India, the UK? From anywhere that isn't China or the US?
There are none. Mistral Small 4 is pareto-competitive in its pricing bracket at $0.15/$0.60, at worst it's second to Gemma 4 26B A4B. The above countries have never had a model that is even close to being so.
This particular Mistral Medium looks to be uncompetitive at that pricing. I'm surprised it's so expensive given its size. Wonder if we'll see other providers offer it for cheaper.
but that doesn't mean Mistral has never produced anything useful.
You mentioned "pareto-competitive", and EXAONE certainly was that. The statement that the "above countries have never had a model that is even close to being so" is simply too broad.
You're talking about EXAONE 4.5 33B? Gemma 4 31B was released 1 week earlier and blows it out of the water. Which point in time/model size are you possibly talking about? The original K-EXAONE in January?
More than anything the availability speaks for itself. If it was indeed pareto competitive, all dozens of model providers would be doing their best to offer it for serverless inference. They don't. There's maybe one that does. Do you think a lot of companies wouldn't prefer a Korean model over a Chinese one? In this case, the market speaks. Go talk to people who run business based on putting billions or trillions of tokens through open weights models. And how much time they put into optimization of model selection to save money and latency. And ask why none of them are using EXAONE models. It's not because we're not aware of their existence. There's also reason to believe they've been benchmaxxing more than Chinese models, btw. Have you done the vibecheck?
I wish they were strong, I hope that in the future, they are. More diversity is better. So far they have not yet been a serious option at any point.
Without Google’s funding its not obvious i DeepMind would have went anywhere.
Unless the moved to US for funding while keeping a back office in the UK.
It’s strange to expect anything significant to come out from Europe when VCs there are either very risk averse and/or don’t have enough cash to begin with. It’s not like government or EU funding can replace that since its almost always wasted or missdirected
It’s a company containing such remarkable talent that I’m sure they would not have run into significant issues raising capital on international markets.
It’s not like VCs are only allowed to invest in companies in their own country.
Although the Manus decision might change things for AI, Singapore-washing is quite rampant among Chinese companies, so I wouldn't call this place of origin an alternative market.
A few months ago China was being criticized left and right on how somehow it was not able to compete, and once DeepSeek showed up then all the hatred shifted onto how China was actually competing but exploring unfair competitive advantages.
Funny how that works.
Also, aren't the likes of OpenAI burning through over $2 of investment for each $1 of revenue?
2 businesses working to get money from the same customers in the same field is competition. Kellogs is competing with store brand cereal. People are choosing to use these Chinese AI apis because they are good enough for some workflows and cheaper. If they didn't exist, the money would go to the frontier labs. There is no world where this would not be defined as competition.
I find it funny how people don't realize the technical achievements and papers coming out of deepseek or Alibaba. They are making this whole AI thing sustainable and cheap and available to do at home. That's the future. I should be able to run my own harness and model and never bother with openai or anthropic at all.
Qwen3.6 runs on a single GPU and beats claudes sonnet. In benchmarks and real world tests from humans. Kimi is awesome but most people won't be able to host it themselves.
A lot of people are slowly realizing the moat of 1T closed source models is gone as of the last few weeks. It's going to change the industry. April was a huge month for open models, it'll be curious to see if that continues.
This Mistral submission is another nail in the coffin.
> China is not competing, it is distilling US models.
I think you should check your notes. The likes of Kimi K2 thinking shows up as high as the second best general purpose model currently in existence. It seems they compete just fine.
If you believe "distilling" is all it takes to put together a model at the top of any synthetic benchmark then I wonder what you would have to say about all US models that greatly underperform in comparison and still manage to be used extensively in professional settings.
But your argument is an emotional one and not rarional, isn't it?
According to benchmarks which are gamed to the extreme these days. Trusting them blindly isn’t exactly rational either. They don’t necessarily translate that well to real world tasks
It’s obviously not “distilling” as such but there are reasons why Chinnese models are consistently several months behind OpenAI/Antropic
I don't mind Chinese but US under Trump is a fascist state based on ethnic and theological grounds pretty much or soon would be if electorate doesn't decide otherwise.
China and rest of the world has sane leadership that aren't mentally retarted.
I would rather support Chinese tech companies then American ones who write manifestos, bomb children, praise wwii Germany, can't stay online, are publicly making weapons for wars I don't support, etc.
Chinese AI companies are just trying to make money. They are also publicly contributing to forward the field. We all get to decide, but claiming deepseek is involved in genocide is beyond a stretch. Claiming anthropic and chatgpt are... Actually not so much given the president was threatening it and enabling it with an ally...
In lockstep over the past month, a subset of people, un-labelable, unprompted, share this train of thought:
- Mythos wasn't released widely.
- But Anthropic shared info on it and said it was dangerous.
- Anthropic is a company.
- Companies like money.
- Therefore Mythos is marketing hype.
- Remember GPT-2? That also wasn't released. They said it was dangerous.
- But, GPT-3, GPT-4, GPT-5, etc. were released.
- Therefore GPT-2 being dangerous was marketing hype.
I've seen the idea that GPT-2 not being released was marketing hype at least 6 times since Mythos was shared.
It's Not Even Wrong, in the Pauli sense: they weren't selling anything! They weren't raising funding! What were they marketing!?
And there's a lot more elided from history, ex. they didn't have an API yet.
GPT-3 was released, a year or two later, and did have an API. But, no one used it, it wasn't good enough yet. And they did treat it as dangerous, it was wildly over-the-top manually monitored for anything resembling not-intended-use. I got permanently suspended for using the word "twink"
I'm using mistral-medium-2508 for some text transformation operations. It's giving me better results than mistral-large for my use cases.
Looking forward to testing this new model, although I'm not sure if it's really meant at replacing the previous medium model since it's a lot more expensive and presented more as a coding / agentic model (mistral-medium-2508 was priced $0.4/$2 per 1M tokens, mistral-medium-3.5 is $1,5/$7.5).
So that has the alias "mistral-medium-latest", but the official ID is "mistral-medium-2508" which suggests it's the model they released in August 2025.
But... that 1777479384 timestamp decodes to Wednesday, April 29, 2026 at 04:16:24 PM UTC
Wow. I get that "how well can it make SVGs" isn't the (or a) gold standard for how useful a model is or isn't, but the fact the Gemma 4 26B A4B I'm running locally can blow it out of the water doesn't give me high confidence for the model. Maybe an unfair comparison, but...
Drawing SVGs isn't something I really care about either, and I think it's still to "qualitatively compare" e.g. "Opus's pelican vs GPT's pelican vs GLM's pelican" or whatever the kids are doing.
But what stands out to me is that it's barely able to draw a "recognizable" pelican at all. The Devstral 2 model even looks slightly better, though maybe I'm splitting hairs: https://simonwillison.net/2025/Dec/9/
It's so bad I don't want to spend the 18 EUR just to test it for a month. It can't even create an SVG of the facebook logo. There should be plenty of examples of that around.
I'm curios: are you doing a real apples to apples comparison, or are you running a harness that already curates prompts? There's a far and wide margin how any of these models respond based on already loaded context. Most models are pretty much hot garbage until their context is curated appropiately.
I just copied and pasted each prompt as specified by Mashimo and simonw into a chat interface, using a 4-bit Unsloth quantization of Gemma 4 26B, with the default sampler settings recommended by Google, and a system prompt of "You are a helpful assistant". The results are miles ahead of what the Mistral model output.
I've gotten a lot of use out of Mistral models, and I imagine this model is pretty good at other things, but it really feels like a 128B parameter dense model should be at least a little better than this.
This release Mistral really reminds you of the gap between the frontier labs and everyone else.
Pre-agent, there wasn't always an obvious difference between models. Various models had their charms. Nowadays, I don't want to entertain anything less than the frontier models. The difference in capability is enormous and choosing anything less has a real cost in terms of productivity.
I've been a big fan of the smaller labs like Mistral and especially Cohere but it's been a while since I've been excited by a release by either company.
That said, I'm using mistral voxtral realtime daily – it's great.
Coding has always been the main real-world business usecase since day one. There has been no point since the very first public availability of GPT 3.5 in November 2022, that it wasn't.
A lot of us have been agentic coding since almost 2 years ago, mid-2024. I have. The productivity gap of "best vs 2nd vs 3rd best model" was biggest back then and has slowly been shrinking ever since.
> Pre-agent, there wasn't always an obvious difference between models. Various models had their charms. Nowadays, I don't want to entertain anything less than the frontier models. The difference in capability is enormous and choosing anything less has a real cost in terms of productivity.
It's just apples to oranges.
There is not a clear, across the board, winner on non-agentic tasks between Gemini, ChatGPT, and Claude - the simple chatbot interface.
But Claude Code is substantially better than Codex which itself is notably better than Gemini-cli.
In this vein, it should not be surprising that Claude Code is way better than non-frontier models for agentic coding... It's substantially better than other frontier models at specialized agentic tasks.
I’ve been comparing Claude Code and Codex extensively side by side over the past couple of weeks with my favorite prompting framework superpowers…
From my perspective, Claude Code is decidedly not better than Codex. They’re slightly different and work better together. I would have no issues dropping CC entirely and using codex 100%.
If you’re working off of “defaults”, in other words no custom prompting, Claude Code does perform a lot better out of the box. I think this matters, but if you’re a professional software developer, I’d make the case that you should be owning your tools and moving beyond the baked in prompts.
> Pre-agent, there wasn't always an obvious difference between models. Various models had their charms. Nowadays, I don't want to entertain anything less than the frontier models.
This is a very naive and misguided opinion. In most tasks, including complex coding tasks, you can hardly tell the difference between a frontier model and something like GPT4.1. You need to really focus on areas such as context window, tool calling and specific aspects of reasoning steps to start noticing differences. To make matters worse, frontier models are taking a brute force approach to results which ends up making them far more expensive to run, both in terms of what shows up on your invoice and how much more you have to wait to get any resemblance of output.
And I won't even go into the topic or local models.
The Vibe CLI is really bad on Windows, sure they don’t officially support it, so can’t blame them, but a FYI for anyone wanting to try it. It can’t get find and replace right.
This is a very interesting strategy that might pay off. This model is a very good option for enterprise self host. I would argue a lot of companies are VRAM constrained rather than compute constrained. You could fit 4-5 running instances on one H100 cluster where you can only fit 1-2 Kimi K2 or GLM5.
KV size correlates with attention parameters which are a subset of active parameters. So a typical MoE model will have way lower KV size than a dense model of equal total parameter count.
Given what Vibe already did in the previous versions with codestral-v2, that's great news. Keep up the good work ! I don't want to depend on the world's two hungry superpowers.
I like the idea of Mistral, but the last time I evaluated Mistral Vibe it was really nice for $15/month but not as effective as Gemini Plus with AntiGravity and gemini-cli. I am currently running Gemini Ultra on a 3 month 'special deal' and AntiGravity with Opus 4.7 tokens is pretty much fantastic.
That said, when I stop spending money on Gemini Ultra, I will give Mistral Vibe another 1-month test.
I like the entire business model and vibe of Mistral so much more than OpenAI/Anthropic/Google but I also have stuff to get done. I am curious if Mistral Vibe for $15/month is a stable business model (i.e., can they make a profit).
How do you feel about the responsiveness of gemini-cli? I tried it on a paid plan and the 10-minute hang-ups (per step, not the whole plan execution) really break the illusion of performance gains, unless you run it in the background and do something else in the meantime. It's more noticeable when Americans are awake.
With most OSS releases being MoEs, and modern GPUs optimized for MoEs, can somebody with knowledge of the topic explain or speculate why Mistral might have opted for a dense model?
The advantage to a dense model like this Mistral one is that it is as smart as a much larger MoE model so it can fit on less GPUs. The tradeoff is that it is much slower since it has to read 100% of its weights for every token, MoE models typically only read about a tenth (though sparsity levels vary).
I'm rooting for Mistral. It seems they are making a big bet that smaller models will win over larger ones and I can see it happening. I was running some simple chat and tool-calling benchmarks for small models and Mistral Small 4 performed well for it's price ($.15/$.60). Seeing this today got me excited, benchmarks seems solid compared to models much larger, but it's priced higher than Haiku, 5.4 mini, and all the the Chinese models it's comparing itself too. It's not even winning those benches either, just being competitive with them, which is great, those models are 5x+ the size, but they are also 1/2 the price. Hard to be excited about that.
One thing in particular I was disappointed in was its bad explanations when asking about French grammar. It made multiple mistakes and the other models got it right, even Qwen 3.6 27b!
There's a good chance that they'll catch up. The "AI race" is a race to the bottom, with the leaders blowing huge wads of cash on capabilities that get replicated months later by the competition at a fraction of the cost.
The only benefit of leading is mindshare. OpenAI is doubling down on that, by investing in communication companies. That's their pathetic attempt at a "moat".
They catch up by distilling frontier models. They will eventually figure out how to prevent that from happening. No one has any interest in investing tens of billions if the product can be copied and sold for less.
Difficult to say, this information is not really public. That said, those investors include EU agencies and European multinational companies and governments. It’s not as flashy as the ridiculous sums OpenAI is getting but it should be enough to keep them going for a while.
They also have a different business model. They are selling their expertise to fine tune and adapt their models to on-premises computers (which they can help you build) to handle confidential data and information. I would not be surprised that the revenue they get from normal people is negligible in comparison.
I believe they'll get profitable sooner than their frontier competition. Their operating costs seem to peanuts compared to the providers they're compared to most often while having the local advantage of not being Chinese nor American.
GP is stating that the second best in the field, the Chinese, is so far behind the best in the field, GPT 5.5, that it is not even worth testing anything else.
I am not following this obsession with SOTA and benchmark rankings
I have been using DeepSeek and GLMnmodels with OpenCode and Codex and Claudr side by side.
I have not found the Chinese models lacking. I enjoy for coding and like to maintain full control of my codebade and deeply care about the GOF patterns. So I am very stringent in terms of what I want the LLM to code and how to code.
So from my perspective, they are all about the same.
That I agree with, but for more complex autonomous changes the differences are considerable. However, it seems that most models will reach the saturation time in which they will be useful for almost everything and the difference will be in more and more niche and specialized tasks.
They did not stop using it due to contamination. They said it's flawed and indirectly said anthropics results were impossible. It's very possible they are sore losers
Gemma has been better for us at EU languages than mistral (for comparable sized models) :/ so ... dunno. What mistral does well and others are lagging behind is deploying on prem with their engineers and know-how, offering tuned models for your tasks and finetuning on your own data. (I expect google to start offering this next)
It's sad that despite their strength in this for onprem, they're so behind on this in the cloud. No publicly available cloud SFT at all. Meanwhile Google has been offering that for years - though remains to be seen if they will for Gemini 3 when GA.
And on top of it a range of providers like Fireworks and so on that offer it for Chinese models. This seems such an obvious thing for Mistral to offer.
GLM 5.1 is an excellent model, but even at Q4 you're looking at ~400GB. Kimi K2.5 is really good too, and at Q4 quantization you're looking at almost ~600GB.
This model? You can run it at Q4 with 70GB of VRAM. This is approaching consumer level territory (you can get a Mac Studio with 128GB of RAM for ~3500 USD).
For the Claude-pilled people, I don't know if you only run Opus but when I was on the Pro plan Sonnet was already extremely capable. This beats the latest Sonnet while running locally, without anyone charging you extra for having HERMES.md in your repo, or locking you out of your account on a whim.
Mistral has never been competitive at the frontier, but maybe that is not what we need from them. Having Pareto models that get you 80% of the frontier at 20% of the cost/size sounds really good to me.
reply