Skip to main content
Back to Prompts Library Reference response formats for the Chat Completions API. OpenAI and Azure OpenAI use identical response structures.

Tool Calling with Streaming

Streamed response when the model invokes a function call. Tokens are delivered incrementally; the final chunk includes usage data.

Tool Calling Response

Non-streaming response when the model invokes a function call. The full response is returned at once with finish_reason: "function_call".

Streaming Response

Standard streaming response for text generation. Tokens are delivered incrementally; the final chunk includes usage data and finish_reason: "stop".

Regular Response

Standard non-streaming response. The full content is returned in a single JSON object with finish_reason: "stop".