Moonshot launches Kimi K3.1 Pro with cheaper API and a focus on programming

Model updates

Moonshot AI has unveiled the Kimi K3.1 Pro, an upgraded version of the K3 model designed for programming, 3D creation, and financial research. The model is now available on the company’s hosted services, with the API offering lower rates than its predecessor.

API pricing

The platform charges $2 per million input tokens and $10 per million output tokens for the K3.1 Pro, compared to $3 and $15, respectively, for the K3—a reduction of around 33% for each rate. The Highspeed variant, which Moonshot claims is two to three times faster, costs $4 for input and $20 for output per million tokens. Pricing is based on usage and is not included in Kimi or Kimi Code subscriptions. The rates do not include applicable taxes.

Available features

The documented context limit is 1,048,576 tokens, the unit the system uses to divide processed content. This number indicates how much material can be inserted into the context, but does not guarantee that all responses are accurate. The full model weights have not yet been released, and using the hosted service differs from downloading the files for execution on one’s own infrastructure.

Future plans

Moonshot has announced plans for an IPO in Hong Kong. The company has pledged to release the full weights of the K3.1 Pro by November 4, 2026.