Kimi K3.1 Pro

As of October 26, 2026, Kimi K3.1 Pro by Moonshot AI costs $2.00 per 1M input tokens and $10.00 per 1M output tokens, with a 1M-token context window.

Kimi K3.1 Pro is Moonshot AI's October 23, 2026 release, a post-trained version of Kimi K3 rather than a new pretraining run, and it arrives a third cheaper than the model it builds on: $2 per million input tokens and $10 output against $3 and $15 for K3, with cache hits at $0.20. The API ids are kimi-k3.1-pro and kimi-k3.1-pro-highspeed, the second a faster serving lane Moonshot describes as 2 to 3 times the speed at double the token rates. Billing is flat with no tiering by context length, so a 900K token prompt costs the same per token as a short one. The context window is 1,048,576 tokens, max_completion_tokens defaults to 131,072 and can be raised to the full window, and input covers images and video as well as text, with images sent as base64 or an ms:// file reference rather than a public URL. Thinking is always on and reasoning_effort accepts low, high, or max, defaulting to max; temperature, top_p, n, and the penalty parameters are fixed and cannot be changed. Moonshot points it at coding from quick fixes through full project builds (including Unity, Unreal Engine, and Godot), 3D modeling and parametric design with scripting for Blender and CAD tools, financial modeling and research, and general agentic knowledge work. One cost caveat worth pricing in: Artificial Analysis measured K3.1 Pro using about 14 percent more output tokens per task than K3, so the per-task saving is smaller than the per-token cut. Weights are not published yet. Moonshot says the full checkpoint lands by November 4, 2026, which would make it open weight like K3; until then treat it as hosted only. Available on Kimi.com, Kimi Work, Kimi Code, and the Kimi API.