Limits

You control cost and response size on the client side.

Recommended client controls

  • Set max_tokens / max_output_tokens to cap response size.
  • Use stream: true for low-latency UIs.
  • Apply your own rate limiting per end-user.