Kimi K3 Max Tops Open-Weight Models on Agent Arena
Arena.ai ranks Kimi K3 Max first among open-weight models across agent evaluations.
TLDR
Arena.ai's Agent Arena leaderboard places Kimi K3 (Max) from Moonshot AI at the top of open-weight models for confirmed success, praise versus complaint, steerability, and bash recovery. The model records zero tool hallucinations and leads in five of seven domains overall. It also ranks first among open-weight entries in Frontend Code Arena for Brand & Marketing, Data & Analytics, Simulations, and other categories. These positions reflect performance on tasks that require sustained tool selection and outcome improvement in real agent workflows.
Combined views
81.9K
5 Sources, first seen 63d ago