Mistral Large 4
Mistral launched a public preview of Mistral Large 4 (le Chonk), a 1 trillion parameter multimodal model with 49 billion active parameters, with open weights due by the end of October.
Mistral Large 4 is Mistral's largest model so far, up from the 675B Mistral Large 3. It was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own European datacenters. Mistral says it significantly outperforms any open-weight model from the US or Europe and is competitive with the strongest open models globally. The preview API is available on Mistral Studio, and a version with reduced moderation is being red-teamed with cybersecurity partners until the weights are released.
- Date
- Tuesday, 6 October 2026
- Lab
- Mistral AI
- Kind
- model
- Access
- closed API
Figures
| Measure | Value | Measured by |
|---|---|---|
| Total parameters | 1 trillion Mistral's blog says 1 trillion; Mistral's documentation specifies 1.05T | company |
| Active parameters | 49 billion Mistral's blog says 49B; Mistral's documentation lists 52B | company |
| Training hardware | 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own datacenters in Europe | company |
Preview API only for now. Mistral says weights will be released by the end of October (Techmeme/VentureBeat cite October 27). Benchmark claims are Mistral's own.
Sources
This record was checked and corrected against its sources on 6 October 2026. How we check