Europe just flexed. On October 6, 2026, France's Mistral AI unveiled Mistral Large 4 — a 1-trillion-parameter beast nicknamed 'Le Chonk' — and the AI world collectively gasped. Why? Because this isn't just another model. It's Europe's shot at the open-weight crown, and it's built on a mountain of American silicon. That's like showing up to a marathon in borrowed Nikes and claiming you're the fastest runner alive.
Let's break it down. Le Chonk is a Mixture-of-Experts model with 49 billion active parameters. Translation: it's huge, but only a fraction of its brain lights up at once, making it efficient. It's natively multimodal, trained on 3,800 Nvidia chips inside European data centers. Artificial Analysis says this makes France home to the most intelligent model outside the US and China. Not best in the world — best outside the two superpowers. That's the pitch: Europe as the third pole.
But here's the twist. Right now, Le Chonk is API-only, a 'Research Public Preview.' The open weights — the thing everyone actually wants — are promised for the end of October. And the sources can't agree on the date. The Next Web says October 27. Mistral's own Hugging Face page says October 31. That four-day gap? That's where disappointment lives. For a trillion-parameter model, any weight release is a multi-hundred-gigabyte download event. If it slips, the hype curdles fast.
And the attacks are already flying. A Hacker News commenter eviscerated the 'EU-made' branding: 'They're essentially borrowing Chinese open source distillation/innovation off the US frontier running on US/Taiwanese designed chips and calling it EU made.' Ouch. Mistral trained on Nvidia chips, likely fabricated by TSMC in Taiwan. Europe's sovereignty story is running on hardware it doesn't control. That's like declaring independence while your entire economy runs on imported oil.
Then there's the benchmark problem. Mistral claims Le Chonk beats GLM 5.2 from China's Z.ai. But independent benchmarks are still pending. Artificial Analysis does list it as leading in coding and vision — that's meaningful because they're an independent evaluator. But the headline 'beats China' rests on self-reported numbers. Mistral vs. Z.ai. Europe vs. China. The open-weight crown is the prize.
Finally, the license question. When Mistral shipped Large 2 in July 2024, it used a research-only license. Nobody knows what license Le Chonk will carry. If 'open weights' arrives with restrictions, the community will revolt. The 'open weights end of month, who's ready?' framing could curdle into 'who got played?'
So who wins? Mistral, if the weights ship as promised — it owns the 'best open model from US or Europe' title. European enterprises and governments get a compliance asset. Nvidia gets 3,800 chips' worth of validation. Losers? Z.ai and the Chinese open-weight labs, if the benchmarks hold up.
But the real question: can Europe truly claim AI sovereignty when its flagship model runs on American chips? That's the elephant in the room. Trends suggest this debate will define the next phase of AI geopolitics. Keep an eye on October 31. If Le Chonk ships, it could be a watershed moment. If it doesn't, it's just another overhyped promise.