Responses

The responses endpoints create and manage model responses for agentic and multi-turn workloads. Carry the model's reasoning across turns with stateless encrypted reasoning replay or with server-managed state (previous_response_id), alongside tool calling, search grounding, and file inputs.

• Create a response
• Retrieve a response
• Delete a response
• Cancel a response
• Count input tokens
• Schemas

For usage examples, previous_response_id patterns, and guidance on when to use the Responses API versus Chat Completions, see the Responses feature page.