Mistral AI Releases Mistral Large 4 With 1.05T Parameters
Mistral AI launches Mistral Large 4 (Le Chonk), a 1.05 trillion parameter MoE model featuring native image input and a 1M token context window.
Stock photo for illustration only, not from the actual event
- Mistral AI has released Mistral Large 4 (Le Chonk) as a public preview.
- It is a Mixture of Experts model with 1.05 trillion total parameters and 49 billion active parameters.
- Features native image input support and a massive 1 million token context window.
- Trained on 3,800 NVIDIA Grace Blackwell GPUs within Mistral's European datacenters.
Mistral AI has officially announced the public preview release of its latest artificial intelligence model, Mistral Large 4, affectionately nicknamed "Le Chonk." This release marks a significant milestone in the development of advanced foundational models.
The model is built on a Mixture of Experts (MoE) architecture, boasting a massive total of 1.05 trillion parameters while maintaining an efficient 49 billion active parameters during inference, ensuring high performance alongside optimized computational efficiency.
Furthermore, Mistral Large 4 introduces robust multimodal capabilities, supporting native image inputs alongside a staggering 1 million token context window, allowing users to process exceptionally long documents and complex inputs seamlessly.
The combination of a Mixture of Experts architecture and a 1-million-token context window highlights Mistral AI's strategic push to balance computational economy with deep analytical capability, a crucial trend for enterprise-grade AI deployment where processing extensive documentation is essential.
Stock photo for illustration only, not from the actual event
On the infrastructure front, the model was trained using a powerful cluster of 3,800 NVIDIA Grace Blackwell GPUs housed directly within Mistral's proprietary datacenters located in Europe, demonstrating robust independent computing capabilities.
Developers and enterprise users can access the model via API immediately, while the open weights are scheduled for an official public release by the end of October 2026.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment