Last updated: April 5, 2026 · Model Architecture · by Daniel Ashford

What is Reasoning Tokens?

QUICK ANSWER

Hidden thinking tokens that reasoning models generate internally before their visible response.

Definition

Reasoning tokens are tokens generated by reasoning models as part of their internal thought process before producing the final visible response. They represent the model thinking through a problem step by step.

How It Works

Models like OpenAI o3 and Claude with extended thinking generate reasoning tokens automatically. They count toward output token costs and add latency, but significantly improve accuracy on complex reasoning, math, and coding tasks.

Example

For a complex math problem, a reasoning model might generate 500 reasoning tokens working through the algebra, then output a 50-token final answer. You are billed for all 550 tokens.

Related Terms

Chain-of-Thought (CoT)
A prompting technique that asks the model to show its reasoning step by step.
Output Tokens
The tokens the model generates in its response — the most expensive part of API usage.
Latency
How long it takes to receive the first token of a response.

See How Models Compare

Understanding reasoning tokens is important when choosing the right AI model. See how 12 models compare on our leaderboard.

View Leaderboard →Our Methodology
← Browse all 47 glossary terms
DA
Daniel Ashford
Founder & Lead Evaluator · 200+ models evaluated