# Mistral releases open weights for Mistral Large 4 model

published: 2026-10-29T17:05:34.194Z
topic: infrastructure
source: https://siliconangle.com/2026/10/29/mistral-releases-open-weights-for-mistral-large-4-providing-sovereign-control-for-the-companys-le-chonk-ai-model/
url: https://aiagentnews.io/stories/mistral-releases-open-weights-for-mistral-large-4-model

Mistral AI has released open weights for Mistral Large 4, a 1-trillion-parameter model it nicknames Le Chonk, three weeks after a preview. The company says the model is among the strongest open models for agentic coding and can run on a customer's own hardware.

Mistral AI announced that the open weights for Mistral Large 4, called ML4 or Le Chonk, are now available. The model has 1 trillion parameters, with 52 billion active at a time, and is natively multimodal. It was first released in preview three weeks earlier and then received three more weeks of reinforcement training and customer testing.

Mistral says ML4 rivals GLM-5.3 on the Artificial Analysis Coding Agent Index while using 47% fewer tokens per task on average. On Terminal-Bench 4.0, which measures agentic coding in a terminal, its score rose from 28% to 41%, tying GLM-5.3 as the top model for coding agents. The company trained the model on 3,800 Nvidia Grace Blackwell GPUs in its European data centers.

ML4 offers low, medium and high reasoning modes, letting teams set reasoning depth for each task. Open weights allow companies to run the model on their own hardware behind their own firewalls. The release follows the recent launch of Voxtral 3, a family of voice models and voice agents, and Mistral plans more models in that family.

## Why it matters if you build agents

Developers can now self-host a frontier-scale open model for coding agents and keep sensitive data away from outside clouds. The reasoning-mode setting and claimed token savings make cost per task a variable to test. The performance figures are company-reported, so check them against your own evaluations before switching.