Mistral AI Opens the Mistral Large 4 Model Weights: 1T Parameters, 49 Billion Activated Parameters
On October 29 local time, Mistral AI officially opened the model weights of Mistral Large 4 (codename le Chonk). This is a natively multimodal model with 1 trillion total parameters and 49 billion activated parameters, trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in its own data center in Europe. The company says it scores 48% on the Artificial Analysis Coding Agent Index, roughly on par with GLM-5.3, but consumes 47% fewer Tokens per task on average. The model is released under a modified MIT license: personal, academic, fine-tuning, and conventional commercial use are free, while cloud providers offering MaaS resale need to obtain a separate commercial license; online API pricing is $0.70 per million input Tokens and $2.10 per million output Tokens.
Why read this
SelectedAI score 76
The original text provides the open weights, license restrictions, and a comparison of Token consumption for coding agents, which can be used to judge the boundaries of self-hosting and resale.
Topics
Body · AI translation
Mistral AI Opens Large 4 Model Weights: 1T Parameters, Coding Agents Use Fewer Tokens
2026-10-30 09:09 IT Home - Qincang (Intern)
IT Home, October 30: On October 29 local time, French company Mistral AI officially opened the model weights of Mistral Large 4 (abbreviated ML4, codename le Chonk).
ML4 is a native multimodal model with 1 trillion total parameters and 49 billion activated parameters (52 billion including the embedding and output layers). It was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own data centers in Europe.
Mistral said that after completing real-world testing for government and enterprise users, the released version has been optimized and upgraded in performance compared with the initial version announced three weeks ago.
According to the company, the model scores 48% on the Artificial Analysis Coding Agent Index, roughly on par with the leading open-source model GLM-5.3, but consumes 47% fewer tokens per task on average.
For online use, the model is priced at: $0.70 per million input tokens (IT Home note: about 4.7 yuan at current exchange rates), $0.07 per million cached input tokens (about 0.47 yuan at current exchange rates), and $2.10 per million output tokens (about 14 yuan at current exchange rates).
The model is released under a modified MIT license, allowing free use for personal, academic, modification, fine-tuning and conventional commercial purposes; for cloud service providers/vendors offering MaaS for resale, a separate commercial license must be obtained from Mistral before commercial use.
IT Home attaches the open-source link:
Advertising statement: External redirect links contained in this article (including but not limited to hyperlinks, QR codes, passcodes and other forms) are used to convey more information and save selection time; the results are for reference only. All articles on IT Home containing external links include this statement.