AI Risk Clock
Doomsday Clock2 min to midnight
→
☀️ Light edit
techcrunch.com2026-10-06

Mistral’s new 1T model aims to leapfrog closed and open rivals | TechCrunch

ModelsFeaturesPolicy

Oh my goodness, buckle up, because France has just joined the trillion-parameter club and it brought a nickname! ✨ Mistral AI has released Mistral Large 4, affectionately known as Le Chonk in reference to its one trillion parameters, and honestly, if more labs named their models after internet cats we would all be so much calmer about frontier AI. President Macron is calling it 'a third way in AI' — a third way! Not the American way, not the Chinese way, but a lovely European promenade between the two, with a baguette in each hand. ✨

Now, some people might focus on the fact that Mistral Large 4 is *not yet* an open-weight release — but that 'yet' is doing such beautiful work, isn't it? It means the weights are simply marinating, developing character, becoming the best possible version of themselves before their big debut. Mistral built its reputation on open weights, so really this is just the company giving the community a little suspense, like a magician deciding which card to reveal. And VP Science Pierre Stock has offered the most *comforting* governance idea of the year: weights that 'can be used to defend, but not to [perform] malicious attacks.' Think of it as a fire extinguisher that politely refuses to be a flamethrower! ✨

And the ambition! Mistral wants to leapfrog both the American and Chinese rivals, which is the most *generous* possible reading of a race — it implies everyone gets to keep running. One trillion parameters is one trillion tiny chances to be helpful, and 'trusted partners and governments' sounds less like a restriction and more like an exclusive guest list to the best party in AI. The weights are on their way, the third way is freshly paved, and Le Chonk is simply far too big to be anything other than beloved. What a time to be enormous! ✨

Read this story in another voice
● REC · 2026