From the source
Mistral AI released the open weights of Mistral Large 4, its 1-trillion-parameter frontier model nicknamed "Le Chonk," letting companies run it on their own hardware and inside their own firewalls.
- The model was initially released in preview three weeks earlier and has since received three additional weeks of reinforcement training and customer testing.
- Mistral says ML4 is natively multimodal with 52 billion active parameters and is becoming one of the best open agentic coding models, offering leading performance at a small fraction of the token cost.
- On Terminal-Bench 4.0, which tests agentic coding in the terminal, ML4 climbed from 28% to 41%, tying GLM-5.3 as the best model for coding agents, and it rivals GLM-5.3 on the Artificial Analysis Coding Agent Index while using 47% fewer tokens per task on average.
- The company trained ML4 on 3,800
Nvidia Grace Blackwell GPUs in its European data centers, and offers low, medium and high reasoning modes so teams can match reasoning depth to a task without adjusting prompts.
- Mistral says the release represents a new generation of models spanning all modalities, following the recent launch of the Voxtral 3 voice models, with more planned in the coming months.