reasoning token

Reasoning tokens are the tokens a reasoning model generates in the intermediate steps where it breaks down a problem and checks its work. They are distinct from the tokens of the final answer shown to the user, and some services report only their count without revealing their content.

API pricing usually bills them as output tokens. As a result, a model that reasons at length can cost more per task even at the same price per token, and settings such as reasoning effort are used to control how much it reasons.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.