French AI company Mistral has released the weights for its 1 trillion-parameter open-source AI model "Mistral Large 4 (ML4)", TechCrunch reported on Oct. 29 (local time).

Companies and government agencies can download the model and run it themselves. ML4, known as "Le Chonk", was first unveiled early this month.

Open-weight AI developers such as Mistral argue that companies and governments do not need to rely on external AI firms to use frontier AI. TechCrunch reported that the claim could gain more traction in cybersecurity, where users want direct control of data and systems.

Arthur Mensch (아서 멘쉬), Mistral CEO, said, "AI should become a tool to build capabilities, not a means of dominance."

Mistral claims ML4 delivers top-tier performance with only a fraction of the computing and capital that OpenAI and Anthropic are pouring in.

Mistral said ML4's performance is similar to China's open-weight model GLM-5.3. GLM-5.3 has been regarded as a leading open-weight model, particularly for cybersecurity tasks.

Mistral also said ML4 ranked No. 1 among open models in the agent coding category on the Coding Agent Index run by AI benchmark site Artificial Analysis.

Pierre Stock (피에르 스톡), Mistral's vice president of science, said it means companies and governments can use state-of-the-art AI to block cyber attacks. He said companies can install ML4 in their own environments to identify and respond to major security incidents in real time.