Create chat completion
OpenAI-compatible chat completion endpoint. Streaming support is reserved for a future capability.
Authorizations
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Body
"gpt-5.4"
Sampling temperature.
0 <= x <= 2Nucleus sampling probability mass.
0 <= x <= 1Deprecated compatibility alias for max_completion_tokens. APIAny converts it only when the selected model route requires the newer field.
x >= 1Maximum reasoning plus final-answer tokens. Kimi K3 defaults to 131072 when omitted and supports up to 1048576.
1 <= x <= 1048576Kimi K3 reasoning effort. K3 always reasons and currently supports only max.
max Up to 4 stop sequences.
Number of completions to generate.
x >= 1-2 <= x <= 2-2 <= x <= 2Best-effort deterministic sampling seed.
Force JSON object or JSON-schema output, e.g. { "type": "json_object" }.
Tool / function declarations the model may call.
Tool selection: 'auto' | 'none' | 'required' | { type: 'function', function: { name } }.
0 <= x <= 20End-user identifier for abuse monitoring.
If true, partial deltas are streamed as server-sent events (OpenAI chat.completion.chunk), terminated by 'data: [DONE]'.