Moonshot added Kimi K3.1 Pro to its Kimi API on October 23, 2026, and says full model weights will be released by November 4, according to developer Simon Willison. Until then it is available only through the API.
Pricing is $2.00 per million input tokens and $10 per million output tokens. Willison notes that matches GPT-6.1 Sol and is half the price of Claude Opus 5.5. A Highspeed version runs two to three times faster at twice the price, except for cached input tokens, which cost the same.
On independent scoring, Artificial Analysis rated Kimi K3.1 Pro at 55, ahead of Claude Fable 5.1, GPT-6 Astra and Gemini 4 Argon at 53 and behind only Claude Opus 5.5 at 58. Moonshot's own benchmarks put it roughly level with Opus 5.5 and GPT-6 Astra. Willison calls it the first time a soon-to-be open-weight model has scored this highly against the frontier.
Open weights means the trained model files can be downloaded and run on your own machines, a bit like being handed the engine instead of renting the car. The announcement covered here does not state the licence terms, the model size or the hardware needed to run it, and those details decide who can actually use it.
Why this matters for banks and fund issuers
For institutions that cannot send client data to an outside API, a near-frontier model with downloadable weights would widen the set of models that can run on their own servers or a private cloud. The decision still rests on the licence, the hardware bill and the institution's own tests on its own tasks, not on a leaderboard score (see how LabForty sets up local AI around your boundaries). The next checkpoint is November 4, when the weights are due.
