Bullish
Kimi K3 Model Weights Released: 2.8T Parameters and Native Visual Capabilities
2026-07-27 23:18:24
Kimi AI publishes weights and report for K3, a 2.8T-parameter MoE model featuring native visual understanding and 1M-token context support.
Woofun AI reports that Kimi AI has published the model weights and technical documentation for the Kimi K3 system. The release identifies K3 as the developer's most capable iteration to date, characterized by a 2.8-trillion-parameter Mixture of Experts architecture.
The model incorporates native visual understanding capabilities and supports a context window of up to 1 million tokens, marking a significant expansion in processing scope.
WOOFUN AI
Impact Assessment · Quick Read
The open-sourcing of a 2.8T-parameter MoE model lowers barriers for high-end AI development, potentially accelerating innovation in the open-weight sector. Native visual capabilities and massive context windows position K3 as a strong competitor in multimodal AI tasks. This move may pressure other providers to release comparable architectures or enhance their own open-source offerings.
Generated by WOOFUN AI · For reference only, not investment advice
Comments
No comments yet.