When Does Claude Reset? (5-Hour, Weekly, Fable)

Claude resets on rolling windows, not at any clock time. The 5-hour session window resets exactly 5 hours after the first message of your current cycle. The weekly all-models cap resets 7 days after the first message of your current week. On Max plans, the Fable meter sits inside that same 7-day window rather than running a clock of its own. Nothing on Claude resets at midnight, on Sunday, or at any fixed time - the rolling-window math is the most common misconception about Claude's limits, and the one that bites people planning around a Sunday-night reset that never comes.

Update, 2026-09-19: Max plans no longer run a separate weekly limit scoped to Opus. Opus and Sonnet are both current, fully available models, and both draw from the single weekly limit that covers every model. The second weekly meter on Max is scoped to Fable instead: Max includes Fable for up to 50% of the weekly limit, after which Fable keeps running on pay-as-you-go usage credits. Pro has one weekly limit across all models and bills Fable from usage credits from the first message. Checked against a live Claude Max 5x account that used Fable, Opus and Sonnet in the same week: Settings > Usage shows current session, this week, and Fable this week, with no Opus meter.

The short answer

Your next reset time, in your local timezone

If you started a brand new cycle right now, here is when each of Claude's three caps would reset, computed in your local timezone:

The three windows Claude uses

Claude's consumer plans (Pro and Max) run two or three rolling windows simultaneously. Pro has two; Max has all three. The mechanics are identical between tiers - only the budget size differs.

The session window is the one that bites short bursty work - a 4-hour debugging session can chew through it. The weekly all-models is the one that bites Claude Code users on long stretches. The Fable meter bites Max users who lean on Fable, because it is included only up to 50% of the weekly limit before it switches to pay-as-you-go usage credits. Max plans used to run a separate weekly Opus cap on top of the all-models pool; that meter no longer appears. For the underlying mechanics of each cap see how Claude's usage limits really work in 2026.

How rolling windows actually work

A rolling window starts the moment you send your first message of the current cycle and ends exactly N hours or days later. It does not snap to clock boundaries. It does not align with other users' windows. It does not align with calendar days.

Worked example: you send your first message of the week on a Tuesday at 9:42 AM. Your weekly window now runs until the following Tuesday at 9:42 AM. If you hit the weekly cap on Thursday at 4:18 PM, the cap stays locked until next Tuesday at 9:42 AM - not Wednesday morning, not midnight Thursday, not Sunday night, not next Monday. The specific time the next Tuesday at 9:42 AM is when your budget refreshes.

The 5-hour session is the same shape on a shorter timer. First message at 11:08 AM, session window runs until 4:08 PM. Hit the session cap at 3:55 PM and the budget refreshes at 4:08 PM - 13 minutes of waiting. Hit it at 11:20 AM, the budget refreshes at 4:08 PM - nearly 5 hours of waiting.

This is the source of the most common misconception about Claude's limits: people plan around clock-based resets that do not exist. The two questions to anchor on are when did this window start and therefore when does it end. Both numbers are on Settings > Usage on claude.ai in your local timezone.

When the cap will not reset early

Three things people sometimes hope will reset Claude's caps but do not:

• Refreshing the page or signing out and back in. The cap is server-side. The percentage you see is your real percentage, and refreshing only updates the display. Logging out does not pause the rolling window.
• Switching from the web app to Claude Code or the mobile app. All Claude surfaces share the same usage pool on your account. Moving between them does not give you a fresh budget.
• Buying extra usage credits. If you have purchased extra usage on Max, those credits are a separate budget that unlocks after you hit the weekly cap. They do not reset the cap itself - they just give you a parallel budget to draw from while the cap is locked.

Two things that do change the picture:

• Upgrading from Pro to Max. The upgrade gives you Max's larger budget effective immediately, which can pull you back from a cap you just hit on Pro. The window mechanics stay the same; the budget inside them grows.
• Time passing. The only mechanism that resets a Claude cap is the rolling window completing. Nothing accelerates it. The actionable lever is pacing, not waiting smarter.

What peak hours mean for the reset (they don't)

The reset itself is mechanical. 5 hours is 5 hours regardless of how loaded Anthropic's cluster is. Peak hours - roughly 9 AM to 5 PM US Pacific time - do not affect when your cap resets.

What does change during peak hours is the experience inside the cap. Pro users on peak hours see more "Claude is at capacity, try again" errors, slower model responses, and more failed tool calls that need to be retried (and the retry costs you tokens twice for the same task). The cap itself doesn't bend; only the friction inside it does. Running the same work at 6 AM PT or 9 PM PT gives smoother responses for the same budget. For the full tactical playbook see 13 ways to save Claude tokens and make your session last longer.

Where to see your real reset time (not a projection)

The calculator at the top of this article is a projection - "if you started a new window right now, here is when it would reset." For your actual reset times, you need a surface that reads your real cap state from Anthropic's backend. Three options:

• Settings > Usage on claude.ai - the official source of truth. Shows each cap's percentage and the exact next reset time in your local timezone. Per Anthropic's help-center article on Pro usage, this is the authoritative view.
• /usage in Claude Code - the same numbers in your terminal. The fastest path if you live in the CLI.
• A menu-bar tracker like tokn.watch - the same numbers updated every few minutes in your menu bar. No active checking needed; the reset countdown sits in your peripheral vision so you do not have to remember to look. For the full comparison see the complete guide to tracking your Claude usage.

One related quirk: the ChatGPT side does this differently

If you came to this article from a ChatGPT background and expected Claude to work the same way, the comparison is worth a sentence. ChatGPT's caps are also rolling windows, but on different timers (a 5-hour window for Codex and ChatGPT Work on Plus, plus a weekly allowance) and OpenAI does not publish the percentage or reset time anywhere - you find out you have hit a cap by hitting it. Anthropic publishes both. For the full side-by-side see Claude vs ChatGPT limits, side by side in 2026.

Frequently asked questions