Text

Use this section for chat, multi-turn conversations, agents, and structured text output.

​

Endpoint list

• /v1/chat/completions
• /v1/responses

​

When to use which

• Use chat.completions for traditional chat-completion clients
• Use responses when you want a more unified interface for text, structured output, and future multimodal expansion

​

Request notes

• Authenticate with Authorization: Bearer {API_KEY}
• chat.completions is centered around messages
• responses is centered around input, instructions, and response_format

​

Response notes

• chat.completions usually reads from choices[0].message.content
• responses usually reads from output_text or output[]
• Streaming clients should handle SSE chunks incrementally

​

Legacy page

• OpenAI text compatibility page