API overview - OpenPAI Docs

OpenPAI exposes three protocol endpoints at once, so you can choose freely based on your existing code or the upstream model.

Endpoint quick reference

Protocol	Base URL	Auth header	Best for
OpenAI-compatible	`https://openp.ai/v1`	`Authorization: Bearer sk-...`	Existing OpenAI SDK code / most models
Claude Messages	`https://openp.ai/v1/messages`	`x-api-key: sk-...` + `anthropic-version`	Claude thinking / cache_control advanced features
Gemini native	`https://openp.ai/v1beta/models/{model}:...`	`x-goog-api-key: sk-...`	Gemini multimodal, function_declarations

The same sk- key works across all three endpoints, sharing the account quota.

/v1/chat/completions, /v1/embeddings, /v1/images/*, /v1/audio/*, /v1/responses, /v1/realtime

/v1/messages, /v1/messages/count_tokens

:generateContent, :streamGenerateContent, :embedContent

/v1/rerank — Cohere / Jina compatible

/mj/submit/*, /mj/task/{id}/fetch

/suno/submit/music, /suno/fetch/{id}

Protocol: HTTPS, HTTP/2 recommended (curl --http2).
Encoding: Content-Type: application/json; charset=utf-8.
Timeout: server default 600 seconds. Streaming responses keep a long connection, with an idle heartbeat every 15 seconds.
Request body size: default limit 32 MB (after compression).
Response headers:
- X-OpenPAI-Request-Id: the request ID — include it in support tickets to speed up diagnosis.
- X-OpenPAI-Group: the user group this request used.
- X-OpenPAI-Quota-Consumed: the amount charged for this request.
- X-OpenPAI-Remaining-Quota: the account balance after the charge.

Auth header formats and examples for the three protocols.

HTTP statuses, error structure and troubleshooting tips.

RPM / TPM limits and handling 429.