Mistral releases the weights of Le Chonk, its trillion-parameter open model built for coders and cyber defenders

What happened? Mistral AI released the open weights of Mistral Large 4, nicknamed "Le Chonk", on 29 October, three weeks after a public preview. The natively multimodal mixture-of-experts model has about 1.05 trillion parameters, of which around 52 billion are active per task, and a 1M-token context window. Companies and governments can download it from Hugging Face, customise it with Mistral Forge and run it on their own infrastructure.

The coding pitch: Mistral says three more weeks of reinforcement learning made it the best open-weight model for coding. It leads Cognition's private FrontierCode 2.0 benchmark among open models at 19.7%, scores 41% on Terminal-Bench 4.0 (up 13 points from the preview) and ties China's GLM-5.3 on the Artificial Analysis Coding Agent Index at 48% while using 47% fewer tokens per task. It now offers low, medium and high reasoning modes.

Built for defenders: The model scores 51% on the Artificial Analysis Cyber Index, and Mistral says it refuses 96.6% of malicious cyber requests. Booz Allen, Cisco, Cloudflare, CrowdStrike, Elastic, Fortinet, Orange Cyberdefense, Rubrik, Zscaler and the EU cybersecurity agency ENISA are among the organisations testing or backing it. Mistral also says live traffic since the preview showed "a potential for uplift to conduct real attacks", so it is tightening guardrails, and a specialised variant for penetration testing and red-teaming is only available to vetted design partners such as Orange Cyberdefense.

The independent view: Artificial Analysis rates it narrowly the top open-weight model developed outside China, with a score of 26 on its Intelligence Index, but still about 6 points behind DeepSeek V4.1 Flash and GLM-5.3, mainly on coding and agentic knowledge work. The weights ship under Mistral's MEOW licence, which requires a separate commercial licence to run a model-as-a-service business.

Why it matters: "Open source is essential to ensure state-of-the-art AI does not end up concentrated in the hands of a few actors," CEO Arthur MenschDealroom has a profile for this one. Try Dealroom → said. Mensch told the etn. show this is just the beginning, with "bigger cats coming".

Read more: Mistral, Hugging Face, Artificial Analysis, Euronews, TechCrunch, etn.