Moonshot AI releases Kimi K3, 2.8T open-weight MoE
Kimi K3 is a 2.8T-parameter sparse MoE with native vision and a 1M-token context window; weights and technical report released July 27. New Kimi Delta Attention and AttnRes components deliver roughly 2.5x better scaling efficiency than Kimi K2, and it ranks #1 open-weight in the Agent Arena.
15 sources
audnai/penclaw-Kimi-K3.0-abliterated-GGUFhuggingface.co
Together AI announces strategic partnership with Moonshot AI to natively serve Kimi modelstogether.ai
Code Arena now measures fullstack capabilities! View overall rankings across AI models on...x.com
Latest open artifacts (#23): Laguna S2.1, Inkling, & Kimi K3 show the utility of open models on the Pareto frontierinterconnects.ai
Kimi K3 Architecture Notessebastianraschka.com
moonshotai/Kimi-K3simonwillison.net
What Is Moonshot AI’s Kimi K3 Model and Why Is It Making Waves?bloomberg.com
On Kimi K3: Its Capabilities And Related Discontentsthezvi.substack.com
Moonshot AI by email
Get an email when Moonshot AI has news
No news that day, no email.
More stories today
- Qwen Code v0.21.5 nightly adds browser-extension readiness diagnostics
- NVIDIA demos Nemotron 3 Ultra running locally on DGX Station
- Sakana AI, Daiwa Securities joint AI project enters full-scale production
- Multi-datacenter architecture gives Together AI 99.9% uptime
- LLM 0.32 adds reasoning traces, OpenAI Responses, server-side tools