Mistral releases Large 4 weights, with strong coding results but lower overall rankings

What happened

Mistral released the weights for Mistral Large 4, nicknamed Le Chonk, three weeks after its public preview. The text-and-image input model shows competitive coding results, while Artificial Analysis rates it at 26 points overall, behind several Chinese rivals, with generation speed reaching 161 tokens per second. API pricing is $0.70 per million input tokens and $2.10 per million output tokens; its modified MIT license requires a separate license for commercial cloud service providers.

Read next

• More changes in today’s briefingGoogle DeepMind announces Gemini 4 Carbon with cybersecurity and long-horizon capabilities

Reported by 5 sources

The story so far

Mistral AI released improved open weights for Mistral Large 4, also called Le Chonk, with launch-day inference support from NVIDIA and vLLM. Mistral says its license allows free use, modification, fine-tuning and commercialization, but businesses operating Model-as-a-Service offerings must obtain a separate license before commercial use. The Decoder describes that exception more narrowly as covering companies offering Large 4 as a commercial cloud service, and reports that access requires review and contractual restrictions. The multimodal model targets agentic coding and cybersecurity. Artificial Analysis lists 1.05 trillion total parameters and 52 billion active parameters, and says it narrowly leads open-weight models developed outside China. Mistral reports a 48% Artificial Analysis Coding Agent Index score using 47% fewer tokens per task than GLM-5.3, and first place among open models on Cognition’s FrontierCode. Those coding claims have a different scope from the overall Artificial Analysis score of 26 reported by The Decoder, which places it behind several Chinese rivals. NVIDIA announced NVFP4 weights on Hugging Face and an optimized Dynamo configuration for distributed inference. vLLM added support for FP8 and NVFP4 checkpoints with a built-in DSpark drafter, and describes the model as supporting a 1-million-token context.