From the source
Mistral has put the weights of Mistral Large 4, nicknamed "Le Chonk," on Hugging Face three weeks after its public preview, under a modified MIT license.
- The model has about one trillion parameters in total, with 52 billion active per step, and takes text and images as input while outputting text and code.
- Mistral says the model beats its open rivals at coding, and after three more weeks of reinforcement learning its Terminal-Bench 4.0 score rose from 28 to 41 percent, just ahead of GLM-5.3 at 40 percent.
- On Artificial Analysis' Intelligence Index the final version scores 26 points, one point above the preview and above the 23-point median for open models of similar size, but behind Chinese open models such as
DeepSeek V4.1 Flash and GLM-5.3.
- Mistral calls Large 4 the leading open model for cyber defense, yet its own chart shows DeepSeek V4.1 Flash at 57 percent and MiMo-V2.6-Pro at 56 percent ahead of its 51 percent on the Cyber Index.
- Through the API it costs $0.70 per million input tokens and $2.10 per million output tokens, down from the preview list price of $1.36 and $4.18, and companies offering it as a commercial cloud service need a separate license from Mistral.