Moonshot AI’s Nov. 5 release of Kimi K3.1 Pro weights on Hugging Face coincided with a leading open-weight result for its Max version, which ranked 13th overall in an agent benchmark.
Kimi K3.1 Pro (Max) placed first among open-weight models. The evaluation drew on more than 17,500 extended sessions involving agent tasks, submitted by users worldwide. The model also led open-weight entries in work, programming and chat.
Benchmark figures showed +7.9% for Bash Reliability and +6.5% for controllability, both top scores among open models. Those results apply specifically to the Max setting; the weights Moonshot announced were for Kimi K3.1 Pro.
