kimi

Kimi on RunAPI

Use OpenAI-compatible Chat Completions at https://runapi.ai/v1 as the primary protocol.

Primary protocol recipe

Authenticate

Set OPENAI_API_KEY to a RunAPI API key and OPENAI_BASE_URL to https://runapi.ai/v1.

Send request

from openai import OpenAI
client = OpenAI(api_key="YOUR_RUNAPI_TOKEN", base_url="https://runapi.ai/v1")
response = client.chat.completions.create(
    model="kimi-k3",
    messages=[{"role": "user", "content": "Explain this finding."}],
)
print(response.choices[0].message.content)
print(response.usage)

For long output, set stream=True and stream_options={"include_usage": True}; consume through [DONE].

Verify result

Require final assistant content, finish_reason, and authoritative usage. Use the returned final answer rather than raw reasoning content.

Stop boundaries

Correct a rejected shape once using the structured error. Retry transport once only before any response or Usage and when replay is safe. Record a terminal error and stop without changing model or protocol. For kimi-k3 and kimi-k2.7-code, start with text history and final answers; add tools, multimodal input, reasoning controls, cache controls, or continuation only when the current RunAPI contract explicitly verifies the exact shape.

Compatibility protocols

Load compatibility protocols only when an existing client requires Anthropic Messages or Gemini contents.

Supported models

References