Mistral

Major AI research company releasing open-weight models under Apache 2.0. Mistral 3 family includes dense models (14B, 8B, 3B) and Mistral Large 3 with 41B active / 675B total parameters as sparse mixture-of-experts. Champions distributed intelligence through open-source AI model development.

Key Features

• 675B-parameter sparse MoE architecture
• Apache 2.0 licensed open models
• Dense models from 3B to 14B parameters
• Multimodal and multilingual capabilities
• Open-source distributed intelligence focus