The most interesting AI release of the week costs nothing. French AI company Mistral has released the weights of Mistral Large 4 — its trillion-parameter flagship model, nicknamed "Le Chonk" — which means any business or government can now download it and run it on their own systems. The weights are live on Hugging Face and through Mistral's API.
Why cybersecurity is the target
Mistral is pitching this model squarely at cyber defence: run it on-premises to detect, analyse and respond to threats across your code and IT infrastructure, with your data never leaving your building — and with access that no outside provider can switch off. That "sovereign" argument is powerful in security, where buyers are exactly the people most paranoid about sending sensitive data to someone else's cloud.
The performance claims are Mistral's own, so treat them as claims: the company says Large 4 performs comparably to China's highly rated open model GLM-5.3 while using 47% fewer tokens per task, and that it outranks every other open model on agentic coding in Artificial Analysis' Coding Agent Index. Even if you discount the marketing, a trillion-parameter open model in this class is a serious gift to the open-source community.
The bigger fight: open vs closed
Mistral's leadership frames open weights as empowerment — download it, inspect it, own it — against closed labs like OpenAI and Anthropic, whose models you can only rent through an API. Cybersecurity is the sharpest possible battlefield for that argument, because it is the one niche where control over data and systems matters more than raw convenience. Expect every major security vendor to be testing this model this month.
Free tools from this story you can use today
- Hugging Face — download Mistral Large 4's weights and thousands of other open models, free.
- Le Chat by Mistral — Mistral's own AI assistant, free to use in your browser.
- Google AI Studio — another free way to experiment with frontier-class models and prompts.
Source: TechCrunch. Performance figures are Mistral's claims.
