Create response

Opaque OpenAI Responses passthrough. Set stream:true for response.output_text.delta SSE events. Billed in RODI.

Bearer rd_sk_… only (Authorization header). model is required and must resolve to an OpenAI upstream.

Request body parameters

Parameter	Type	Required	Description
model	string	Required	Catalogue id served by an OpenAI upstream (e.g. openai/gpt-4o).
input	string \| array	Required	Prompt text or Responses input items.
stream	boolean	Optional	When true, returns SSE events (response.output_text.delta, …).
instructions	string	Optional	Optional system-level instructions passed through to the model.
max_output_tokens	integer	Optional	Upper bound on generated tokens, reasoning included. Bounds the cost; set it on every call.
temperature	number	Optional	Sampling temperature when supported.

…

…

…

Served for OpenAI models only (for example openai/gpt-4o, openai/gpt-5.1, openai/gpt-5.1-codex). Other models are not available on this endpoint: use /v1/chat/completions.
The body is passed to the provider as-is, apart from model; fields such as instructions, tools or reasoning work as documented by OpenAI. service_tier values priority, fast, ultrafast and scale are not forwarded (the request runs at the standard tier); auto, default and flex are.
background: true and conversation are rejected with 400 unsupported_parameter: use stream: true for long runs.
previous_response_id works only with the id of a response your account created through RodiumAI; any other id returns 404 previous_response_not_found (param: previous_response_id).
Provider web searches (web_search_call output items) are billed per call on top of tokens, see Pricing.
stream: true returns the Responses event stream (response.output_text.delta, then response.completed with usage).
Authenticate with Authorization: Bearer only. Smart aliases and custom models are not supported here.
Set max_output_tokens to bound the cost; reasoning tokens are billed as output.

Streamed deltas → Streaming reference