Kimi K3.1 Pro near Claude Opus 5.5 on Terminal-Bench
Moonshot AI’s open-weight Kimi K3.1 Pro posts self-reported gains over GPT-6 Astra on coding and agent tests, with full weights promised by 4 November.
Moonshot AI launched open-weight Kimi K3.1 Pro, saying it scored 65.7% on Terminal-Bench 4.0 versus 58.2% for GPT-6 Astra and 66.4% for Claude Opus 5.5.
The model has a 1-million-token context window and is a heavily post-trained version of July’s 2.8-trillion-parameter Kimi K3.
A Highspeed variant runs two to three times faster; both are live on Kimi.ai, Kimi Work, Kimi Code, and the API, with full weights due by 4 November.
Moonshot reports leads on several coding and agent benches, but weaker scores on Terminal-Bench Science, ProgramBench, and some diamond exams.
Results are company self-reported at maximum reasoning effort; cybersecurity lags are described as intentional.