Mistral Large 4 enters public preview
Mistral’s new flagship has 1.05 trillion parameters and a million token context window. Downloadable weights are planned for later in October.
Mistral launched Mistral Large 4 in public preview on October 6. The company says developers can try its API through Mistral Studio now, with downloadable model weights planned by the end of October.
The model documentation lists 1.05 trillion total parameters, 49 billion active parameters and a vision encoder with 1.6 billion parameters. Its context window holds up to one million tokens. The multimodal model uses a mixture of experts architecture and supports function calling, structured outputs and agent workflows.
Mistral says it trained the model on 3,800 NVIDIA Grace Blackwell GPUs in its European datacenters, which also serve the preview. The company is targeting coding, cybersecurity, finance and legal work. Its launch performance claims come from the company’s evaluations and cited benchmarks.
Teams can begin evaluating the hosted API now. Those planning to run the model on their own infrastructure still need the promised weights release. The announcement does not specify a license for those weights.
Mistral AI logo via LobeHub under the MIT License. Mistral trademarks belong to Mistral.
[…] examines published results and access details rather than reporting tests run by ByteForward. Our earlier launch report covers the model’s architecture and release […]