Mistral has released the weights for its open model, Mistral Large 4 (ML4), on Hugging Face, describing it as its flagship open frontier system. The model is a 1 trillion-parameter natively multimodal system with 52 billion active parameters. Mistral said it continued training the model for three weeks using reinforcement learning after an earlier preview, while testing it with customers and cybersecurity partners ahead of the wider release. Mistral said the model ranks as the top open model for agentic coding on Artificial Analysis' Coding Agent Index, and improved its score on Terminal-Bench 4.0 from 28% to 41%, putting it level with GLM-5.3 on that benchmark. On the Artificial Analysis benchmark, ML4 uses 47% fewer tokens per task on average than GLM-5.3 while delivering similar results, Mistral said. The model includes three reasoning modes - low, medium and high. Mistral is positioning the release around coding and cybersecurity use cases, particularly for organisations that want to run models on their own systems, and said the model can be deployed on a customer's own infrastructure for cyber defence work. ML4 was trained on 3,800 Nvidia Grace Blackwell graphics processing units in Mistral's European data centres. Pierre Stock, Vice President of Science at Mistral, said the open weight version continues reinforcement learning training and delivers substantial gains with no signs of saturation.