Output tokens
The tokens generated by the model in its response, reasoning, or tool-call output.
Illustrative conversation
- Why did a short final answer still report many output tokens?
- The usage breakdown may include reasoning or structured intermediate output in addition to the visible text.
Definition
Output tokens are the tokens a model generates after receiving input. They can represent visible text, structured data, or tool-call content; reasoning-capable APIs may report reasoning tokens in a separate output-token breakdown. Output limits cap how much can be generated, and longer outputs generally increase latency and billed generation usage.
Reviewed