Paris-based Mistral has released a preview of Mistral Large 4, nicknamed Le Chonk, a 1-trillion-parameter multimodal model it describes as the most powerful open-weight system outside China. Developers can access it now through Mistral’s API. Full model weights are scheduled for release on 27 October after three weeks of real-world testing with developers, cybersecurity firms, and government agencies. The model uses a sparse architecture that activates only about 49 billion parameters per task. It was trained from scratch over two months on roughly 4,000 Nvidia Grace Blackwell GPUs in Mistral’s European data centers. Preliminary scores include 62 percent on the DeepSWE agentic coding benchmark, 67 percent on a finance benchmark, and strong visual-grounding results. The model supports more than 160 languages, including every official EU language. It is the first major model funded by Mistral’s recent €3 billion round.
Paris-based Mistral has released a preview of Mistral Large 4, nicknamed Le Chonk, a 1-trillion-parameter multimodal model it describes as the most powerful open-weight system outside China. Developers can access it now through Mistral’s API. Full model weights are scheduled for release on 27 October after three weeks of real-world testing with developers, cybersecurity firms, and government agencies. The model uses a sparse architecture that activates only about 49 billion parameters per task. It was trained from scratch over two months on roughly 4,000 Nvidia Grace Blackwell GPUs in Mistral’s European data centers. Preliminary scores include 62 percent on the DeepSWE agentic coding benchmark, 67 percent on a finance benchmark, and strong visual-grounding results. The model supports more than 160 languages, including every official EU language. It is the first major model funded by Mistral’s recent €3 billion round.