Earlier quoted context omitted.
Yeah but why use mistral on premises instead of Qwen?
Please don't run Chinese models for KYC operations.
[dead]
Notes from the Mistral AI Now Summit
51–60 of 230 posts
Re: Notes from the Mistral AI Now Summit
#52Oh most prominent eu ai company . Without reading an article predict next, will update after :
1. They give up on building competitive models. It’s time to drink wine not to struggle with competition
2. Because of #1 they will talk a bit about something around llms maybe coding agents , and after start talking about sovereignty.
Re: Notes from the Mistral AI Now Summit
#53Earlier quoted context omitted.
Nobody trying to compete with Google, OpenAI, and Anthropic should be playing the small models / local models game. Foundation model labs should be building very large reasoning models, then leaving it to the community to distill them down. You can't scale a small model up, but you can scale a small model down. I'm convinced the only way we'll have a seat at the table in the future and avoid total runaway takeoff is…
I do think there's a chance open weight models have a bit of a moment with the costs of frontier models growing on business balance sheets. It's unfortunate from my "privacy loving" PoV that it's mostly Chinese models filling the gap. ( the top models on openrouter for instance ). I have used Mistral models out of pure ideology for web agents and the like which aren't doing a lot of heavy lifting.
Antirez’s Deepseek 4 Flash implementation that can run on MacBooks also was a revelation. It runs decently on M5 Max 128GB and it’s pointing out other bottlenecks like prefill speed which will improve.
Re: Notes from the Mistral AI Now Summit
#54Re: Notes from the Mistral AI Now Summit
#55Oh most prominent eu ai company . Without reading an article predict next, will update after : 1. They give up on building competitive models. It’s time to drink wine not to struggle with competition 2. Because of #1 they will talk a bit about something around llms maybe coding agents , and after start talking about sovereignty.
3. They are going to start focusing on B2B implementation and deployment.
See what happened to Aleph Alpha...
Re: Notes from the Mistral AI Now Summit
#56OK, I'm 100% rooting for both Mistral and task focused small models. But Mistral has fall really far behind since 2025Q3. It seems they can't get good reasoning models working at even medium context sizes, which is necessary to be at the table right now. Gemma4 and Qwen3.6 are currently best in the small size; Mistral's "small" model has ~4x the parameter count at 120B and isn't even competing with models a quarter i…
agreed, the next price increase from frontier labs (and the inevitable limits decrease in subscription tiers) will have people thinking real hard about their model providers and that's when mistral should be ready. however, given their recent performance, I realistically don't have my hopes high up.
DeepSeek is both cheaper and better than Mistral.
Re: Notes from the Mistral AI Now Summit
#57Re: Notes from the Mistral AI Now Summit
#58Earlier quoted context omitted.
agreed, the next price increase from frontier labs (and the inevitable limits decrease in subscription tiers) will have people thinking real hard about their model providers and that's when mistral should be ready. however, given their recent performance, I realistically don't have my hopes high up.
DeepSeek is both cheaper and better than Mistral.
Because they distill
Re: Notes from the Mistral AI Now Summit
#59OK, I'm 100% rooting for both Mistral and task focused small models. But Mistral has fall really far behind since 2025Q3. It seems they can't get good reasoning models working at even medium context sizes, which is necessary to be at the table right now. Gemma4 and Qwen3.6 are currently best in the small size; Mistral's "small" model has ~4x the parameter count at 120B and isn't even competing with models a quarter i…
Yeah. I run LLM models locally and for me 22B-32B is the largest I'm willing to invest in trying out.
Even though Mistral 4 has 6B active parameters per token (allowing 3-3.5 per token parameters to be loaded on a 4090), the ~240GB download + storage is pushing the limits of being able to try this out locally, especially if you are downloading and evaluating multiple models.
It also makes it harder for other people to make downstream finetunes like with what happened with the older Mistral/Magistral models.
Re: Notes from the Mistral AI Now Summit
#60Wasn't even aware Mistral was around and I think that just shows you how irrelevant it has become and not a very good sign for EU in general when the best talent are working for American AI companies.
I saw Tibo's tweet a while back and it was basically a legitimate complaint about the extreme taxation he faced back in EU (France I think) and its pretty obvious how much of a hinderance top down centralized regulation is to innovation.
While I welcome competition and independence, nobody can argue with American innovation and its ability to attract the best of the best. Once it takes seat of the AI reigns there is very little chance for other countries to compete, very much similar to semiconductor field and how only a few select countries have the talent and monopoly over its particular supply chain.
It's clear to anyone looking in that whatever EU is doing is not working (not just AI) and will not work as they do not seem flexible or humble enough to steer itself.