How does streaming work on the Model API?

With "stream": true, the reply arrives as server-sent events while it's being written, in the OpenAI chat.completion.chunk format. Each event is a data: line with JSON; the stream ends with data: [DONE].

• Text is in choices[0].delta.content.
• Max and Heavy stream their reasoning as delta.reasoning_content first.
• Tool calls arrive as delta.tool_calls pieces you join together.
• With stream_options: {"include_usage": true} a final chunk carries usage with empty choices.
• If something fails partway: data: {"error": {...}} then data: [DONE].

Streaming doesn't change the price; you pay for the same tokens either way.

Good to know#

• Every call shows in Console › Logs and Usage as Model API, with its price. See what you spent
• Your key's account is responsible for your end users' text under the Abuse policy. What it means for apps
• New accounts get $5 of free credit; top up from $1 to $800 per purchase in Console › Billing › Credits. Billing

Still need help?#

• General support: support@clatterlabs.com or team@clatterlabs.xyz
• Infrastructure questions, or permanent deletion of your data (your whole account or one service): infrastructure@clatterlabs.xyz

We usually reply within the first hour, and always within 24 hours. A full support team is with you.