Mistral 3

Mistral Large 3 is Mistral’s largest model to date, featuring 41B active parameters and 675B total parameters, with a large 256k context window, and offers powerful agentic capabilities.

Combining compact efficiency with multimodal and multilingual capability. Engineered for edge devices, self-hosted systems, and robotics, these models seamlessly blend language, vision, and reasoning into highly efficient architectures