Kimi K3.1 Pro scores near the top of the frontier, open weights promised

Moonshot released Kimi K3.1 Pro through its API on October 23, and the benchmark numbers put it in territory open-weight models have not reached before. Artificial Analysis gave it a 55, ahead of Claude Fable 5.1, GPT-6 Astra, and Gemini 4 Argon, which all sit at 53, and behind only Claude Opus 5.5 at 58. Moonshot's own tests place it roughly level with Opus 5.5 and GPT-6 Astra, winning a few benchmarks and coming close on the rest. Simon Willison, who ran the model through his usual pelican-drawing tests at low, medium, and max reasoning effort, notes that this is the first time a soon-to-be open-weight model has scored this high against the closed frontier.

The weights are not out yet. Moonshot says the full model will be released by November 4. Pricing is $2 per million input tokens and $10 per million output, which matches GPT-6.1 Sol and is half the price of Opus 5.5. A Highspeed variant runs two to three times faster at double the price, though cache-hit input tokens cost the same. The model supports low, medium, and max reasoning levels.

Why it matters

If you have been waiting for an open-weight model you can run yourself that is genuinely close to the best closed models, this is the one to test when the weights land on November 4. Check it on your own tasks before switching, since the headline scores come from Moonshot and a single aggregate index.