Mistral Falls Behind Big Three And Chinese AI Labs Despite Talent And Backing

I get that Mistral has largely given up on building frontier models, as many comments have pointed out, but it absolutely did not start out that way. (& a bet on being a vendor of frontier open weights models is only a good bet as long as Chinese labs are allowed to release them)

@emollick They are not *trying* to build frontier models anymore. Like Cohere, they realized there is a good business in AI integration services dressed in nationalist colors. But it means the govts are not getting what they think they are from these subsidies...

@ramshackle78 Sure, sovereign contractors are fine. But Mistral started as a model maker, and those are much rarer than contractors. They were, in fact, the only EU frontier model maker. It seems significant that they walked away from it.

@emollick They don’t need to because that’s not their business model

Oh Ethan, you need to spend more time in Europe. Mistral are not trying to compete with the big three and the Chinese labs. Instead they work with Europes largest companies on deploying Ai behind the firewall. So so data leakage. They are more like Palantir than OpenAI or Anthropic.

Lately, with European countries distancing themselves from American tech, Mistral has positioned itself as a friendlier, “sovereign” and homegrown alternative

https://techcrunch.com/2026/06/12/mistral-is-rumored-to-be-raising-e3b-at-e20-valuation/

Mistral disproved the idea that Europe lacks AI talent. What it may have proven instead is that talent alone can't overcome a compute and capital gap measured in tens of billions. Frontier AI increasingly looks like a scale business, and that's a much harder problem for Europe to solve.

I find absolutely galling coming from you to call Mistral a failure. That's a complete misunderstanding of what they are trying to do, where they're at, who they are. But also, everywhere else you're preaching 'adoption logic' but here you're talking as if frontier benchmark is all there is. Sovereign AI prime contractor might not do it for X discourse but it matters IRL

@mllichti @emollick Why would any AI company in Europe run towards a business model that OpenAI are themselves running away from?

“Chat is dead,” said one senior OpenAI employee.

https://giftarticle.ft.com/giftarticle/actions/redeem/92f47f77-6436-46ea-a7a8-72ac0183a2f7

@emollick My guess:they were a fruit of French corrupt processes rather than meritocracy. All they could do was copy other open source models. (Source: used to work for the French government.)

i fear Mistral isn’t really trying to win the raw frontier scaling race the same way the others are. they’re aiming for European sovereignty: open-weight models, EU-hosted, customizable for banks/gov/defense who can’t or won’t use US or Chinese hyperscalers.

worth noting Mistral’s CEO has said China isn’t lagging, their open models are already stressing US labs.

(as per https://www.mobileworldlive.com/ai-cloud/mistral-boss-insists-china-is-not-lagging-on-ai/)

@emollick This is being shared as the key problem. Orders of magnitude less compute.

@emollick Since it is almost impossible to raise private funding for tech projects in the EU, Mistral’s future is likely to be a satellite project allowing b the big guys to operate in the EU with a bit of a local footprint.

@mllichti @emollick “There is nothing so useless as doing efficiently that which should not be done at all.” –

Peter Drucker

Once you know Arthur has had lunch with Yann LeCun, then you understand everything.

@emollick This article from April talks about their shift in focus:

https://www.forbes.com/sites/iainmartin/2026/04/16/how-frances-mistral-built-a-14-billion-ai-empire-by-not-being-american/

But then this other one from May indicates they might return toward the frontier race:

https://www.wsj.com/tech/ai/mistral-chases-ai-superintelligence-to-counter-u-s-dominance-b2a44fa1

@emollick Mistral has access to 0.0002 terawatts of AI compute. They’re starved.

Mistral and the frontier labs aren’t competing on the same thing anymore.

Mistral’s models are less powerful for sure (like Macron would say 🙃), but they cover 80% of enterprise use cases in 2026. And they’re laser-focused on business integration now

In a world where EU regulations might force companies to use European/soverain models, that’s actually a great position for them

@emollick I do not think they need to have a SOTA model to be successful. They can provide a "good enough" model that checks the (thanks to Trump) very important "made in EU" box, and they are also focusing on providing end-to-end solutions, not just access to a model.

@emollick Maybe the Mistral guys get off work at 3pm everyday, and don’t work on 996 schedules like the Chinese and American AI lab bros do.

@emollick Compute is one critical factor. We're limited on EU deployments because there isn't enough. This is my first hand experience.

EU must build compute or even a small but flexible country could attract that investment and own that future.

@emollick Maybe death by a thousand cuts?

Backward induction likely convinced many to prefer to work elsewhere if possible? Same with VC etc.

Unless you have decent odds to be front, lots of small decisions go against you