Output token
Last verified
Output tokens are what the model produces. They cost more than input tokens — commonly 3–5× — because generation is inherently sequential: each token depends on the one before it, so it cannot be parallelised the way reading a prompt can.
Two things make output cost larger than it looks:
- Reasoning tokens are output tokens. Models that reason before answering bill that internal work as output, and you never see it in the response. A short answer can carry a large output bill.
max_tokenscaps a response, not a run. An agent making 200 capped calls is still expensive.
When estimating cost, weight output heavily. A workload that generates a lot of text is priced very differently from one that reads a lot of text.