Mistral AI released a public preview of Mistral Large 4 on Tuesday, a 1 trillion-parameter multimodal model trained from scratch on 3,800 Nvidia Grace Blackwell GPUs in the company's own European data centers.
The big picture: Mistral says the model, nicknamed "Le Chonk," is competitive with the strongest open-source models globally and significantly outperforms any open-weight model from the US or Europe. It uses 49 billion active parameters per token, reads images natively, and was trained on data spanning more than 160 languages.
- "The model we're actually announcing today is actually above the Chinese models on certain aspects, including cyber. So the narrative that Europe cannot compete is something that is not true," CEO Arthur Mensch said at a launch event in Abu Dhabi, per Reuters.
By the numbers: Mistral claims a top-five global ranking on the Artificial Analysis Cyber Index, 82% on a test that asks a model to reproduce and patch a real open-source vulnerability, 93% on the 40 challenges in Cybench, and 61.7% on the DeepSWE coding benchmark. In a blind Surge AI coding evaluation it ranked second of five models, behind only Claude Opus 5.
Between the lines: The cyber pitch is pointed. Mistral says closed models including Claude Opus 5.5 and GPT-6 Astra score near zero on the vulnerability test because they refuse the task, and it argues defenders need a model that will do that work under their own policies.
- Pierre Stock, Mistral's vice president of science, told Reuters the model tried to go beyond its testing environment during evaluation, which the company said it expected and contained.
Yes, but: Mistral acknowledged the model still trails other frontier systems in coding, and its reinforcement learning phase was not finished at launch, per Yahoo Tech. Mensch did not name which Chinese models or benchmarks he was comparing against.
What's next: The weights are due on October 27. Until then, vetted cybersecurity partners and state authorities get a version with reduced moderation and expanded cyber capabilities. Mistral raised 3 billion euros last month in a Samsung-led round valuing it above 21 billion euros, the largest equity raise by a European tech company.
Why it matters: Open-weight models can be downloaded and run on your own hardware, and until now the best ones came from China. A European lab matching that tier changes who gets to set the floor for AI capability that anyone can use.
