The best open model outside China did not come from the U.S. It came from Paris. Mistral launched Large 4, nicknamed Le Chonk: one trillion parameters, 49 billion active, factory multimodal. On aggregated benchmarks, it is the strongest open-weights model ever built in the U.S. or Europe—in other words, outside China. Until now, the world’s most capable open models were Chinese. The API is already live. Weights will be released at the end of October. Trained from scratch in Europe on 3,800 Grace Blackwell GPUs within Mistral’s own data centers. Operates under European law. The strike against closed models: in visual grounding, it outperforms closed frontier models. In cybersecurity, on the Artificial Analysis Cyber Index, it ranks in the global top 5 and leads all open models outside China. In a test to reproduce and patch a real-world vulnerability, it scored 82%—the highest of any model. Claude Opus 5.5 and GPT-6 Astra scored near zero, because they refused the task. Defending software often begins by proving a vulnerability exists. Closed models freeze. Open models, in the hands of companies, do not. The battle now is this: the U.S. and Europe in closed models. China in open models. Paris has entered the game with weight.