What Are Reasoning Tokens (and Why They Blow Up Bills)?
Reasoning models produce a large amount of internal thinking that you never see but still pay for as output tokens. That hidden volume is why a reasoning model can cost many times more than its visible answer implies.
Hidden thinking is billed output
Before answering, reasoning models generate chains of intermediate tokens. Those count as output at the full output rate, so a short final answer can sit on top of thousands of billed thinking tokens.
When reasoning is worth it
Reserve reasoning models for genuinely hard problems where accuracy pays for itself. For routine tasks, a standard model is far cheaper — routing by difficulty usually beats sending everything to a reasoning model.
Calculators for this topic
Frequently asked questions
Do I pay for reasoning tokens?+
Yes — hidden reasoning is billed as output tokens even though you do not see it in the response.
How do I control reasoning cost?+
Use reasoning models only for hard tasks, and lower the reasoning effort where the provider allows it.