Does extended thinking count toward the Claude usage limit?
Updated
Yes. Anthropic's help center says a higher effort level uses more tokens, 'so you'll reach your usage limits faster', and the Claude Code docs state 'Thinking tokens are billed as output tokens, and the default budget can be tens of thousands of tokens per request'. That output draws from one pool across claude.ai, Claude Code and Claude Desktop. Lower the effort with /effort or /model; thinking cannot be turned off on Fable 5.1 or Opus 5.
What the official docs say
Higher effort means more thorough responses, but they take longer and use more tokens, so you'll reach your usage limits faster.
Quoted from Claude Help Center: Change the model, effort, and thinking settings Thinking tokens are billed as output tokens, and the default budget can be tens of thousands of tokens per request depending on the model.
Quoted from Claude Code docs: Manage costs effectively Tokens Claude uses while thinking (billed as output tokens)
Quoted from Claude Platform docs: Steering thinking and cost your usage of all different Claude product surfaces (claude.ai, Claude Code, Claude Desktop) counts towards the same usage limit.
Quoted from Claude Help Center: How do usage and length limits work? Thinking cannot be turned off in Claude when using Claude Fable 5.1 or Claude Opus 5.
Quoted from Claude Help Center: Change the model, effort, and thinking settings You can't turn off thinking on Fable models.
Quoted from Claude Code docs: Manage costs effectively
How it is counted
Thinking is output, and output is metered. The Claude Platform pricing page lists 'Tokens Claude uses while thinking (billed as output tokens)', and the Claude Code cost guide says 'Thinking tokens are billed as output tokens, and the default budget can be tens of thousands of tokens per request depending on the model.'
Effort is the dial. The help center says 'Higher effort means more thorough responses, but they take longer and use more tokens, so you'll reach your usage limits faster.' Usage is also affected by which model you chat with and the effort level you selected, and 'your usage of all different Claude product surfaces (claude.ai, Claude Code, Claude Desktop) counts towards the same usage limit.'
You can turn it down but not always off. In Claude Code, lower the effort level with /effort or in /model, or disable thinking in /config, but 'You can't turn off thinking on Fable models.' The help center says the same for the consumer apps: 'Thinking cannot be turned off in Claude when using Claude Fable 5.1 or Claude Opus 5.'
| Setting | Counted toward the session / weekly limit? | Source |
|---|---|---|
| Effort level: higher | Yes. More tokens, "so you'll reach your usage limits faster" | [1] |
| Thinking tokens | Yes. Billed as output tokens; the default budget can be tens of thousands per request | [3] [4] |
| Surface: claude.ai, Claude Code, Claude Desktop | All count toward the same usage limit | [2] |
| Turning thinking off | Possible via /config in Claude Code, but not on Fable models; not on Fable 5.1 or Opus 5 in the apps | [1] [4] |
FAQ
Can I turn thinking off to save usage?
Not on every model. The help center says 'Thinking cannot be turned off in Claude when using Claude Fable 5.1 or Claude Opus 5', and the Claude Code docs say 'You can't turn off thinking on Fable models.' On other models you can disable it in /config or lower the effort with /effort.
Do claude.ai and Claude Code share the same limit?
Yes. Anthropic says 'your usage of all different Claude product surfaces (claude.ai, Claude Code, Claude Desktop) counts towards the same usage limit.' Thinking tokens spent in one surface come out of the pool the others use.
Are thinking tokens cheaper than normal output?
No. The Claude Platform pricing page lists 'Tokens Claude uses while thinking (billed as output tokens)', and the Claude Code docs repeat that thinking tokens are billed as output tokens with a default budget that 'can be tens of thousands of tokens per request'.
How do I lower the effort level in Claude Code?
The Claude Code cost guide says you can lower the effort level with /effort or in /model, or disable thinking in /config where the model allows it. Run /usage to see the session and weekly bars afterwards.
Sources
- Claude Help Center: Change the model, effort, and thinking settingsverified 2026-09-17
- Claude Help Center: How do usage and length limits work?verified 2026-09-17
- Claude Platform docs: Steering thinking and costverified 2026-09-17
- Claude Code docs: Manage costs effectivelyverified 2026-09-17
Want the next Claude reset on your phone?