Le Chonk: Built for Coders, Trusted by Defenders.
mistral ·
Mistral has released the full weights of Mistral Large 4, also known as Le Chonk, a 1-trillion-parameter multimodal model designed for coding and cybersecurity. The release follows a successful public preview, during which reinforcement learning significantly improved performance in code generation and defensive security tasks. It currently leads open-weight benchmarks on platforms like FrontierCode and Terminal-Bench while offering superior token efficiency compared to competitors. To balance cost and performance, the model introduces three distinct reasoning modes: Low, Medium, and High. A key feature is its robust safety guardrails, which maintain high refusal rates for malicious cyber requests while effectively assisting with legitimate vulnerability detection and patching. The model supports Sovereign AI initiatives by allowing organizations to self-host, customize, and deploy it on their own infrastructure without relying on external APIs. Major partners including Microsoft, NVIDIA, and various cybersecurity firms have endorsed the model for its transparency and control capabilities. Mistral Large 4 is available under a modified MIT license, with specific commercial terms for Model-as-a-Service offerings.
mistral