Output token
Also known as Completion token, Output tokens
Definition
An output token is a unit generated by a language model in its response. Output tokens include visible text and, depending on the API, structured fields or reasoning content returned for the application to process.
Output length is a design choice
The model may generate explanations, code, JSON fields, or tool arguments. A maximum output token setting limits the response but does not guarantee a useful stopping point. Clear instructions and structured schemas can reduce unnecessary generation.
Cost and latency
Providers commonly charge output tokens at a different rate from input tokens. Longer responses also take longer to stream and can delay downstream agent steps. Track output length with quality and completion rather than optimizing for brevity alone.
Find the right budget
Set a budget by task type, then inspect truncation, retries, and user corrections. Weave's token intelligence makes those tradeoffs visible across models and routes.
For agent workflows, count output tokens from intermediate steps as well as the final answer. Those hidden costs can explain why a seemingly short user response still consumes a large budget.
How this relates to Weave
Weave relates output token volume to route, completion time, and task outcome. That helps teams spot verbose responses, runaway generations, and cases where a shorter answer performs just as well.
Explore Token intelligence