Mistral Bets ‘Le Chonk’ Can Give Europe an Open-AI Edge

Mistral Bets ‘Le Chonk’ Can Give Europe an Open-AI Edge
Mistral on Tuesday put a trillion-parameter marker down in the global AI race, unveiling Mistral Large 4—nicknamed “Le Chonk”—as it seeks to prove Europe can still compete with Chinese open-model developers and America’s closed-model giants.

The French company says ML4 is especially suited to cybersecurity, coding, finance, manufacturing and multimodal work. It was trained over two months on 4,000 Nvidia Grace Blackwell GPUs in Mistral’s European data centers—a leaner compute footprint than the company says many Chinese and closed-source competitors use. Its release follows Mistral’s €3 billion Series D fundraise in September, which valued the company at €21 billion.

The immediate rollout, however, is deliberately limited. ML4 is available in preview to developers, cybersecurity leaders and state authorities; the public can access it through a guarded endpoint rather than downloadable weights. Mistral says those weights should arrive at the end of October, after safety testing.

That caution sits at the heart of the company’s pitch. Pierre Stock, Mistral’s vice-president of science, said the lab would work with “trusted partners and governments” so the weights can be used “to defend, but not to perform malicious attacks.” He also argues that open-weight systems are easier to audit—a direct rebuttal to the idea that security requires keeping powerful models closed.

Mistral’s claims remain aspirational in one crucial respect: benchmark results are still pending, and the company acknowledges the model trails the broader frontier in some coding tasks. Still, it says ML4 will be the strongest open-weight model built outside China by a “substantial margin.” The coming weight release will show whether Le Chonk is a genuine third way—or merely a well-funded contender in a race still led elsewhere.

https://foxvector.com