Kimi releases K3.1 Pro API and a faster Highspeed version

Kimi has launched API access to K3.1 Pro and its Highspeed variant. The company says Highspeed offers equivalent intelligence while delivering roughly two to three times the speed of the standard version. That performance comparison is a company claim.

K3.1 Pro supports visual input and can handle up to one million tokens of context. Kimi describes it as the first release in its K3.1 family and says it improves coding, creation of 3D content, and financial analysis and modeling.

The release follows K3, which remains listed in the company’s API catalog. Developers can select three reasoning levels for K3.1 Pro, although its reasoning mode stays enabled. The API also supports tool use and structured responses.

Access requires an account top-up of at least $1, with usage limits tied to cumulative funding. Developers can begin integrating either variant through the API. Kimi cautions that its web search functionality is being updated and is not currently recommended for production use.