Reasoning Tokens
Thinking content that a large model generates internally before producing its final answer. It is typically not visible to the user or is billed separately, and is charged at the output rate.
Timeline (last 90 days)
Reasoning tokens are the internal thinking content the model generates before giving its final answer, usually billed at the output rate, and can account for an extremely high share of total Agent output (77% of the bill in the example).
A study published on OpenReview argues that the reasoning tokens generated by LLMs are fundamentally unrelated to human reasoning.
Context window inflation caused by intermediate tokens or conflicts with compression mechanisms are an entirely different class of problems from the semantic validity of reasoning tokens.
All Facts (3)
Reasoning tokens are the internal thinking content the model generates before giving its final answer, usually billed at the output rate, and can account for an extremely high share of total Agent output (77% of the bill in the example).
50%UnverifiedA study published on OpenReview argues that the reasoning tokens generated by LLMs are fundamentally unrelated to human reasoning.
50%UnverifiedContext window inflation caused by intermediate tokens or conflicts with compression mechanisms are an entirely different class of problems from the semantic validity of reasoning tokens.
50%