French AI company Mistral has released the weights of Mistral Large 4, a trillion-parameter model nicknamed "Le Chonk", on Hugging Face and through its API. Businesses and governments can download it and run it on their own infrastructure. The company is pitching the model at cybersecurity work.
Mistral says the model performs comparably to the Chinese open-weight model GLM-5.3 while using 47% fewer tokens per task, and that it ranks above all other open models on Artificial Analysis' Coding Agent Index. It also claims state-of-the-art results with far less compute and capital than closed labs use. The article does not name the license terms.
VP of Science Pierre Stock said open weights mean access "can't be cut off" and that advanced defensive capabilities should reach everyone quickly, since attackers may also use AI. The release raises pressure on closed-model providers such as OpenAI and Anthropic.