Capabilities

TextFunction CallingStructured Output

Notes

Kimi's flagship model for long-horizon coding and end-to-end knowledge work. Always-on reasoning (reasoning_effort, currently only 'max'). Context caching, tool_choice, and dynamic tool loading supported. Note: web_search is being upgraded — not recommended for now.

Official Details

Long Context1,048,576 tokens (1M) context window

Input (cache hit)¥2.00/M

Reasoning effortAlways-on, max only

Other Models from This Provider