French AI company Mistral has released the weights for its trillion-parameter open model Mistral Large 4 (ML4), nicknamed “Le Chonk,” with a sharp focus on cybersecurity. Businesses and governments can now download the model and deploy it themselves — no reliance on a closed API.
Mistral AI (Credits: TechCrunch)
Mistral Large 4 “Le Chonk” Trillion-Parameter Model Weights Released
The model, first introduced earlier this month, is positioned squarely against closed labs like OpenAI and Anthropic. Mistral’s argument: organizations should not have to depend on outside AI companies for frontier capabilities — especially in cyber defense, where control over data and systems matters most. As CEO Arthur Mensch put it, AI should be “a tool for empowerment, not dominance.”
Mistral claims ML4 achieves state-of-the-art performance at “a fraction of the compute and capital committed by closed labs,” and that it is comparable to China’s GLM-5.3 — previously considered one of the best open-weight models for cyber work — while using 47% fewer tokens per task. The company also says the model outranks all other open models on agentic coding tasks in Artificial Analysis’ Coding Agent Index.
Why Open Weights Matter for Cyber Defense
Pierre Stock, Mistral’s VP of Science, told TechCrunch that open weights mean organizations can defend themselves with frontier AI that can’t be cut off — and can be deployed on-prem inside their own infrastructure. The model can “identify, coordinate, and respond in real time to major security incidents,” he said, and releasing it openly matters because malicious actors may already be weaponizing their own AI.
“Amid growing cyber threats, ML4 brings state-of-the-art sovereign cyber defense capabilities to enterprises and governments globally,” Mistral said, noting the weights are available on Hugging Face and via its API.
Also Read: Mistral’s Voxtral 3 Brings Speech-to-Text, TTS and a Voice Agent SDK in One Package
Also Read: NVIDIA Releases Nemotron 3.5 Transcribe: Open 30B MoE Model for 40-Language Transcription
Source: TechCrunch AI — read the full story there.
