Kimi K3: an upgrade for open-weight models
AI summary
Moonshot AI released its flagship model Kimi K3 on July 16, a 2.8T-parameter MoE architecture whose weights will be open-sourced on July 27. K3 ranks second on the Vals AI index and third on the Artificial Analysis intelligence index (behind only Claude Fable and GPT-5.6 Sol Max, at a lower price), and ranks first on the Frontend Code Arena, making it the strongest open-weight model to date. The model is based on two architectural updates, Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), and expands MoE sparsity (activating 16 of 896 experts); combined with the Stable LatentMoE framework, overall scaling efficiency is about 2.5 times higher than Kimi K2.
Key Takeaways
- Moonshot AI released its flagship model Kimi K3 on July 16,a 2.
- 8T-parameter MoE architecture whose weights will be open-sourced on July 27.
- K3 ranks second on the Vals AI index and third on the Artificial Analysis intelligence index…
- 6 Sol Max,at a lower price),and ranks first on the Frontend Code Arena,making it the strongest…
- The model is based on two architectural updates,Kimi Delta Attention (KDA) and Attention Residuals…
Quick Facts
- Published
- Updated
- Category
- Opinion
- AI score
- 67
Kimi K3 的开源权重发布让开源模型与前沿闭源模型的差距进一步缩小,文章结合技术细节与行业格局,梳理了效率优势、开源博弈与生态影响,是理解当前开源模型竞争态势的一篇必要阅读。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Kimi K3: an upgrade for open-weight models. July 20, 2026. https://new.nat6.net/en/news/cmrtglz5p32w9bitlhu4cc999/