Does thinking use more allowance?

Reasoning effort affects usage. The same question can consume different amounts with another model or effort setting.

Sources reviewed:

Reasoning effort affects usage. The same question can consume different amounts with another model or effort setting.

Codex

Reasoning, context and task complexity affect usage. Visible reply length is not the whole workload.

Claude

Higher effort uses more tokens. Match effort to the task rather than always choosing the maximum.

Compare fairly

Keep the task, model and context similar. One experiment cannot establish a universal saving.

Sources & status

OpenAI · Codex usage ↗

Claude · Usage and length limits ↗

Back to the tracker →

Copy the address below