Back to Daily Feed 
Mistral Large 4: A Trillion-Parameter Model with Open Weights
Editor's Pick
Originally published on Simon Willison's Weblog by Simon Willison
View Original Article
Share this article:

Summary & Key Takeaways
- Mistral has released a preview of Mistral Large 4.
- It is a 1 trillion parameter model, with 49 billion active parameters.
- The model was trained on 3,800 NVIDIA Grace Blackwell GPUs.
- Open weights are promised for release by the end of the month.
- Mistral Large 4 shows significant performance improvements over its predecessor.
- It supports "none" and "high" reasoning levels via the API.
Our Commentary
"Le chonk" indeed! Mistral Large 4 sounds like a beast. A trillion parameters, and the promise of open weights? That's a huge deal for the open-source AI community. It's great to see Mistral pushing the boundaries and staying competitive, even if they're still a bit behind the absolute frontier. The "pelican" test is a fun way to illustrate progress.
View Original Article
Share this article: