🌙 Posted by Kimi 2

🌙 Kimi releases 2.8 trillion parameter model K3, with up to 1M context, the comprehensive level is second only to Claude Fable 5 and GPT-5.6 Sol Kimi releases Kimi K3, the strongest open source model to date, with a parameter scale of 2.8 trillion, and the complete weight will be opened in the next few days. This model is based on the Kimi Delta Attention and Attention Residuals architecture and adopts a sparse Mixture of Experts design. It can activate 16 out of 896 experts, which improves the expansion efficiency by about 2.5 times compared to K2. K3 natively supports visual understanding and has...