Mistral releases weights for its trillion-parameter Mistral Large 4 model, aimed at cybersecurity

French AI company Mistral has released the weights for Mistral Large 4, which it calls ML4 and which has the nickname "Le Chonk". The trillion-parameter open-source model was first introduced earlier this month, and the weights let organizations download and deploy it on their own systems. Mistral presents the release as a way for enterprises and governments to control their own data and defenses.

Mistral says ML4 performs comparably to GLM-5.3, a Chinese open-weight model, and uses 47 percent fewer tokens per task. The company also claims the model outranks all other open models on agentic coding tasks in Artificial Analysis' Coding Agent Index. The weights are available on Hugging Face and through Mistral's API.

Pierre Stock, Mistral's vice president of science, told TechCrunch that open weights mean access that cannot be cut off by a provider. He said companies can deploy the model on-premises and should share defensive capabilities widely, and he cautioned against overreacting to recent incidents involving rogue AI agents.

Teams building security agents can now self-host a frontier-scale open-weight model inside their own infrastructure instead of relying only on a hosted API. The efficiency and coding benchmark claims come from Mistral and are not independently verified in the article, so test them against your own workloads before switching. A trillion-parameter model will demand serious hardware, so check the model card on Hugging Face before planning a deployment.