They’ve built performance, enterprise utility, privacy, sovereignty, open innovation and strategic partnerships into their core story. It's quite a list. The models are opensource, Voxtral outperforms Whisper in terms of accuracy.
The GPT-OSS-120B release was pretty decent and you could run it on vLLM, Ollama and a bunch of other stuff on day one, despite MXFP4, are you not entertained? I mean, it's even close to GPT-5 mini in some benchmarks: https://llm-stats.com/
As for the Chinese models, yes, there are quite a few good ones.
For programming and development, my current daily driver is the Qwen3 Coder 480B model: https://qwen3lm.com/
Personally I think Claude still has the best results, but Qwen3 is loosely in the same ballpark and Cerebras inference is measured in thousands of tokens per second, in addition to giving me 24M tokens per day for 50 bucks a month in total. That was enough to get me to switch over.
Either way, happy to see what the future holds for Mistral, it's cool to have EU options too! Either way, more competition prevents complacency and stagnation, and should be a good thing for everyone.
What's "serious" exactly? Codex is open source, is software, can be run with open/downloadable models/weights.
In my testing using Gemini, Claude Code, Codex, Qwen Code and AMP side-by-side for every prompt for the last two weeks, Codex seems the best of all of them so far.
Yeah, I initially thought so too, but since they used "models" later, I assumed they knew the difference and really meant "software".
> recent GPT-OSS is not competitive with other open weights models
Yeah, heard that a lot from people who haven't run GPT-OSS themselves too, but as someone who been playing with it since launch, and compared it to the alternatives since then, saying it isn't even competitive is a serious signal they don't know what they're talking about.
There are concerns besides spying if you really don't trust the source of an open model. One is that the training incorporates a bias (added data or data omission) that might not be immediately apparent but can affect you in a critical situation. Another is vendor lock-in, if you end up depending on specifics of the model that make it harder to swap later.
It goes for all models though if you are looking at the values argument that original commenter made -- western values are probably more aligned than authoritarian governments - even if you do have your concerns about western companies. At least thats my read on the situation.
yeah, but try to convince a board or legal about it for a company that is not software first, for that they have to understand how it works. we have "chinese" AI blocked at work, even through i use self hosted models for myself at home hacking on my own stuff.
Good luck convincing others of this. I know it's true, you know it's true, but I've met plenty of otherwise reasonable people who just wouldn't listen to any arguments, they already knew better.
It's theoretically possible that your model will work OK except for code generation for security-relevant applications it will introduce subtle pre-designed bugs. Or if used for screening CVs it will prioritize PRC agents through some keyword in hobbies. Or it could promise a bribe to an office worker when asked about some critical infastructure :)
Sending data back could be as simple as responding with embedded image urls that reference external server.
You are totally right EU commissioner, Http://chinese.imgdb.com/password/to/eu/grid/is/swordfish/funnycat.png
Of course theoretically lots of things are possible with probabilistic systems. There is no difference with open source, openweight, chinese, french or american llms. You dont give unfettered web access to any models (locally served or otherwise) that can consume critical company data. The risk is unacceptable, even if the models are from trusted providers. If you use markdown to see formatted text that may contain critical data and your reader connects to the web, you have a serious security hole, unrelated to the risks of the LLM.
Of course, you want to limit that with training and proper procedures. But one of the obvious precautions is to use a service designed and controlled by a trusted partner.
Having the local LLM process sensitive data is a desirable usecase and more trustworthy than using a “trusted partner” [0]. As long as your LLM tooling does not exit your own premises, you can be technically safe. But yes, dont then click at random links. Maybe it is generally safer to not trust the origin of the local LLM, because it reduces the chance of mistakes of this type ;-)
[0] Trust is a complicated concept and I took poetic license to be brief. It is hard to verify the full tooling pipeline, and it would be great if indeed there existed mathematically verifiable “trusted partners”. A large company with enough paranoia can bring the expertise in house. A startup will rely on common public tooling and their own security reviews. I dont think it is wise to share the deepest darkest secrets with ourside entities, because the potential liability could destroy a company, whereas a local system, disconnected from the web, is technically within the circle of trust. Think of a finance company with a long term strategy that hasnt unfolded yet, a hardware company designing new chips, a pharma company and their lead molecules prior to patent submission, any company that has found the secret sauce to succeed where others failed—-none of these should be using trusted partners in favor of local LLM from untrusted origins IMHO. Perhaps the best of both worlds is to locally deploy models from trusted origins and have the ability to finetune their weights, but the practical processing gap between current chinese and non-chinese models is notable.
Maybe it can not spy on you but models can be totally (e.g. politically) biased depending on the country of origin. Try to ask european-, us- or china-trained models about "Tiananmen Massacre" and compare the answers. Or consider Trump's recent decisions to get rid of "woke" AI models.
Classic problem: "Who do you love more: mum or dad?" ;) Surely it's naive thinking but as the EU citizen I feel like I've got a little more influence on "European censorship" than on any other. I suppose that ASML feels the same way
Agreed. Also, companies tend to prefer having someone else bound by a contract run their AI services. That way they are safe from scandals, by having a scapegoat, and do not spend time doing something orthogonal to their expertise.
Mistral's best models are actually not open-source, and the ones that are open are not particularly competitive with other open-source models these days. Their highest ranked open model on LMArena[1] (mistral-small-2506) ranks below: Qwen3, various DeepSeek models, Kimi K2, GLM 4.5, Gemma, GPT OSS, etc.
All those things you listed as part of that story pretty much apply to any open model, so it's kinda a shite list if you want to be differentiated.
That’s true, but not very relevant. Mistral is not in the business of selling their free models. What they are doing for large companies is building datacenters and providing their proprietary models trained on proprietary and confidential internal knowledge and fine-tuned for specific tasks. No sane European organisation would let a Chinese company do this, and American ones are less and less appealing. There is a significant amount of money to be made there and they don’t need to hop on the AGI hype train. They "just" need to provide fast and competent specialised models.
It's very relevant if any other EU firm can take open models (regardless of provenance) and fine tune them in the same way. Mistral really needs to be producing at-or-near SOTA models for them to be differentiated at all, and they are not.
Not even then. You need to compare the end products, which are not the open weight models.
I don’t care whether the LLM can have "PhD level thoughts" (lol) or is able to code golf like a Facebook engineer. It needs to be able to do its task (so all the infrastructure around the model matters just as much as the model itself) efficiently (so small models have an advantage). There are billions of weights in general-purpose models that are irrelevant for specialised uses.
The way to go is efficient models adapted to their task. It’s exactly the same thing as for industrial robots. Geeks get excited every now and then about humanoid robots, but in the real life we don’t need robots to stand on two legs or our LLM to cite Shakespeare.
> They’ve built performance, enterprise utility, privacy, sovereignty, open innovation and strategic partnerships into their core story.
This has to be a buzzwordiedest sentence i've ever read. what is 'enterprise utility' and how does mistral have that more than any of the other open models ?
Can I stop you right here? Whisper is a few years old and it wasn't the best model for a long time. There are like 10 models that are smaller and faster and outperform both of them.
> There are like 10 models that are smaller and faster and outperform both of them.
As someone who is currently relying on Whisper for some things, what models are those exactly? I still haven't found anything that is accurate as Whisper (large), are those models just faster or also as accurate/more accurate?
Yes for parakeet, but only comparing benchmark results for canary. Whisper also has severe hallucinations on silence and noise and WhisperX helps a lot, it adds voice activity detection i.e. a model to detect when someone speaks, to filter the input before running whisper. https://github.com/m-bain/whisperX
All sounds like classic marketing/positioning angles for an indiehacker bootstrapped saas tool.
Problem is Mistral needs more than $10K MRR, and isn't going to make it by carving off a small niche when each model costs 10s of Billions to train and run. Europe has no solution to the energy problem long term unfortunately, and is actively trying to make it worse.
I'm 100% certain some giant industrial companies in the EU will sign a huge contract with Mistral to give their employees "EU approved" AI.
But I'm also 100% certain these employees will just use chatgpt or any of the other frontier models in actual day-to-day reality. Europeans aren't dumb and don't want to be fed inferior slop in the name of abstract emotional vibes.
Europe has more nuclear than the US currently (in GW and even more by percentage of grid) and is building more currently and has more in serious planning.
From your phrasing I assume you don't believe in renewables so what energy problem solution are you referring to?
I think it's Renaissance Fusion (which is still in the EU, but is not Wendelstein 7-X) that has the solution, but it is as stellerator.
The only iffy thing are those little ceramic balls full of lead that they talk about letting float inside the lithium, but I suppose they lithium flow might be slow.
I don't see how Renaissance Fusion's proposed machine can fail to work.
Is there any source you could reference. Really interested.
It would not surprise me, why would they build from scratch, every LLM is a "fork" of gpt. Did they not come up with the mixture of expert idea though ?
The US equivalent of Mistral is Nous Research [0]. Also there would be no Mistral without Llama and it seems like everyone forgot that their LLMs derived from Meta.
For every 'Mistral' in the EU, there's 3 or 5 of them in the US.
And everyone forgets that electricity was invented (mostly) by Europeans, but so what? Everything comes from something, doesn't make any place inherently better for continuing to inventing more breakthroughs, it's just people in a place after all.
There is no AI company like Mistral.