skip to content
Requests & turns

Cloud models

Requests & turns

One request is one metered turn. The gateway holds no conversation state.


Turns

A chat is a growing messages array the client sends whole, every time. A ten-turn conversation is ten requests, each metered on what it carried in and out.

Agent loops

A tool call comes back, your client appends the result, sends the transcript again. Each round trip is a metered turn. /v1/responses serves that loop natively for Responses-wire tools and is just as stateless.

Switching models mid-conversation

Name a different id on the next request with the same transcript. This is how escalation works: run the cheap model until an answer disappoints, resend the same messages to a frontier id.

Streaming, parameters, time bounds

Defined once, in the API reference.