Model
1 entry · December 2025
A 406-billion-parameter mixture-of-experts model with 32 billion active parameters and a 256,000-token context window, released in Think and Instruct variants.
Models & capabilities · Open weights & ecosystem