When does the Claude usage limit reset?
Five hours after the first message of the session, and the weekly cap at a fixed time each week that Anthropic assigns to your account. Anthropic's Max plan page puts the first half in one line: "Your session-based usage limit will reset every five hours." The clock runs from your first message, not your last, so a session opened at 2 PM refills at 7 PM whether you sent one prompt or forty, and the wall itself tells you the time ("Your limit will reset at 7 PM"). Settings > Usage shows the same thing as two progress bars, one for the session and one for the week. The numbers behind the bars moved on 22 September 2026: with Opus 5.5, Anthropic said it is "increasing five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans", without publishing the new size, and added a rate limit reset "which you can now save and use whenever you choose."
We read the usage pages of five assistants on 27 September 2026, and the five-hour clock turns out to be the industry's clock rather than a Claude quirk. Google's Gemini help page: "Your limit refreshes every 5 hours until you reach your weekly limit." OpenAI's ChatGPT Business page sells its Premium seat on having "no 5-hour usage limit", which is the only place OpenAI names the window at all. The weekly cap is what separates Claude and Gemini from the rest, and it is the one that hurts: a heavy Monday can cost you Friday. So before you wait, look at which bar filled. A full session bar is a five-hour problem. A full weekly bar is a calendar problem, and the fixes further down are for that.
How do Claude's limits compare with ChatGPT, Gemini, Perplexity and Grok?
Claude has the strictest published structure of the five, a session window and a weekly cap, and it is also the most transparent about it: none of the five publishes a message count for its paid plan, and only Claude and Gemini publish both clocks. Everything in the table is from the vendor's own pricing or help page, read on 27 September 2026, for one individual paid plan each.
| Assistant and plan | What is metered | Refill window | Longer cap | The vendor's own route past it |
|---|---|---|---|---|
| Claude Pro, $20 a month | Tokens: conversation length, files, model, effort, tools | Every five hours from the session's first message | Weekly, at a fixed reset time assigned to your account | Usage credits at API rates, a free limit reset when one is offered, or Max at $100 or $200 |
| ChatGPT Plus | Text chat "Unlimited*" ("Usage must be reasonable"); uploads, images, voice, deep research and Codex are "Expanded" with "Limits apply" | Not published for Plus; OpenAI names a "5-hour usage limit" on Business Standard seats | None published | Pro at "5x more usage", or Business workspace credits |
| Gemini, Google AI Pro | "Compute-based usage limits", 4x the free allowance | "Refreshes every 5 hours" | "Until you reach your weekly limit" | Keep talking to Flash-Lite, or AI Ultra at 5x or 20x AI Pro |
| Perplexity Pro | Pro Search count ("a high volume of daily searches"); Best mode has no quota | Not published (Perplexity says "a high volume of daily searches") | None published; advanced models "may be limited during weeks of especially heavy usage" | Best mode, or Perplexity Max |
| Grok, SuperGrok | Not published; x.ai sells SuperGrok as "higher limits" | Not published (Grok's own X account said 30 queries every 2 hours in July 2025) | None published | SuperGrok Heavy |
| Whizi Pro, $29.99 a month | Credits per message, fixed by model: Claude Sonnet 5 is 10 of a 2,000 credit month | 60 messages an hour, on every plan | The UTC calendar month | Powerhouse at 8,000 credits, or wait for the month to turn |
Two rows carry footnotes. ChatGPT's unlimited text chat is starred with "Usage must be reasonable and comply with our policies", and its "Limits apply" on Plus covers uploads, images, voice, deep research and Codex without a number anywhere on the page. Grok publishes nothing: the only dated figure comes from Grok's own X account in July 2025, so read it as a hint rather than a policy. ChatGPT says I've reached my limit does this walk for OpenAI's own wording, and Gemini usage limit reached does it for Google's six limits and when each one resets.
What does "usage limit" mean on Claude?
A usage limit on Claude is a token budget for a window of time, not a count of messages. Anthropic's help center calls it your "conversation budget" and says usage is "affected by several factors, including the length and complexity of your conversations, the features you use, which Claude model you're chatting with, and the effort level you've selected." Two people who each send twenty messages can have opposite afternoons.
Four things drain the budget faster than sending more messages. A long thread, because the whole conversation is context for every new turn; Anthropic's Claude Code docs put it plainly: "a one-line question in a session that has been open all day still draws usage for the whole conversation." Files, for the same reason, unless they sit in a Project, where Anthropic says "cached portions count less against your limits than new content." Model choice: Opus spends the budget faster than Sonnet, and Fable faster still (Anthropic: Fable models "use them faster than other Claude models"). And effort and tools: "Higher effort uses more tokens", and "Tools and connectors are token-intensive." The person who hits the wall after four messages usually attached a long PDF to an Opus chat on high effort; the person who never hits it asks short questions in fresh chats on Sonnet.
The other wall, "maximum length for this conversation", is not a usage limit. That one is the context window, up to 1M tokens on the newest models on paid plans, and waiting does nothing for it. Ask for a summary of the thread, paste it into a new chat, and carry on.
Which fix gets you working again fastest?
Read the wall first, because the fastest fix depends on its wording. A limit named after a model ("You've hit your Opus limit") clears the moment you switch the chat to Sonnet or Haiku. A session or weekly limit does not: Anthropic's Claude Code docs say those are "shared across all models, so the developer can't restore access by switching models", and for those the choices are a limit reset, usage credits, or a second app. Three routes, then, and the bar you are looking at picks between them. A capacity message is a different wall. If Claude says it's overloaded instead of giving you a reset time, read Claude overloaded: that error isn't your usage limit, and switching models is the first fix.
Route one is to wait, but wait correctly. Open Settings > Usage, read the reset time, and check the Resets section first: Anthropic now hands out free limit resets, and its help page says "Limit resets are given occasionally to eligible plans, and set your usage limits back to full when you choose to use one." A reset restores either the session or the week, depending on the offer, you cannot undo it once used, and it applies across web, mobile and Claude Code because "Your usage limits are shared across your Claude account." If there is no reset on offer and the session bar is the full one, a five-hour wait started with your first message, so it is often shorter than it feels.
Route two is to change what you spend, which helps before the wall and after a model-specific one. Switch the chat to Sonnet or Haiku, turn effort down, start a fresh chat with a pasted summary instead of continuing a hundred-turn thread, and move the reference files into a Project so they are cached. If the session or the week is gone, Anthropic sells the way past it: Pro, Max 5x and Max 20x subscribers can turn on usage credits, "billed at standard API rates", and buy them in bundles at a discount ($50 of usage for $45, $250 for $200, $1,000 for $700, up to $2,000 a month). Credits only start charging after the plan's included limit is spent, so turning them on costs nothing until you actually cross the line.
Route three is another app that reaches the same Claude models on a different meter. Whizi, which publishes this page, charges a fixed number of credits per message whatever the message holds: Claude Sonnet 5 is 10 credits, so Pro's 2,000 credits a month are 200 Sonnet messages if you spend them all there, and a ten page attachment costs the same 10 as a one line question (how credits work). You can import your Claude export so the thread comes with you. Where it loses: there is no unlimited tier here either, credits refill only when the UTC month turns over, every plan is capped at 60 messages an hour, Claude Opus needs Powerhouse at $49.99 a month, Starter has no Claude at all, and nothing here replaces Claude Code. It is a way to finish today's work, not a way to stop counting.
Why do I still have usage limits with Claude Pro?
Claude Pro is a multiplier on the free allowance rather than an unlimited plan: Anthropic's pricing page promises "at least 5x more usage per 5-hour session than Free", for $20 a month or $17 a month on the annual plan, and the same five-hour and weekly clocks apply. Max is the same multiplier again, "5x or 20x more usage than Pro per 5-hour session", at $100 or $200 a month. Every step up the ladder widens the window; none removes it. Claude Pro vs Claude Max works out who actually needs the wider one.
Two things make Pro feel smaller than 5x. Fable 5 and 5.1 are outside Pro's allowance altogether: on Pro they "run on pay-as-you-go usage credits rather than your plan's usage limits", so a Fable chat with credits switched off stops at once, and the same chat with credits on is billing you by the token. And the weekly cap counts everything on the account, claude.ai, the desktop app, mobile and Claude Code together, so an afternoon of Claude Code on Monday is missing from the chat window on Thursday.
How do you hit the Claude limit less often?
Start a new chat for each task and keep reference files in a Project; those two habits remove most of the re-reading that empties the budget. Anthropic's own best-practices page adds three more: group related questions in one message, "Provide clear, detailed instructions or questions in each message" so Claude does not need a clarifying round, and on paid plans let Claude search past chats and memory instead of pasting the same context again. Then pick the model by the job: Haiku for reformatting, Sonnet for most writing and code, Opus only where the reasoning is the point, and effort set low for routine turns.
Watch the weekly bar on Tuesday, not Friday. The session refills up to five times a day; the week refills once, at your assigned time, and a Max plan gives you the same shape with a bigger number. If the weekly bar is past half by midweek in ordinary, well organised use, the fix is a wider plan or a second meter, not tighter prompts.
- Read which bar filled in Settings > Usage: the session refills five hours after your first message, the week refills once at your assigned time
- Use a free limit reset if the Resets section offers one; it cannot be undone, so spend it when a deadline is real
- If the wall names a model (an Opus or Sonnet limit), switch the chat to another family and keep going; a session or weekly limit is shared across every model, so switching does nothing
- Start a fresh chat with a pasted summary instead of continuing a long thread, and keep files in a Project so they are cached
- Turn on usage credits (API rates, $50 of usage for $45 in a bundle) only if the weekly bar is the one that is full
- Keep a second meter for the days the week runs out: another Claude-capable app charges per message, not per five hours
Frequently asked questions
Is there a limit to how long you can use Claude?
Yes, on every plan. Claude meters a token budget per five-hour session and, on paid plans, a second budget per week that resets at a fixed time assigned to your account. Pro is at least 5x the free session allowance, Max 5x is five times Pro and Max 20x is twenty times Pro, and past any of them you can wait, use a free limit reset if one is offered, or turn on usage credits at API rates.
How do you skip the Claude usage limit?
You cannot skip it, but you can stop paying for it in time. The routes Anthropic itself offers are a free limit reset in Settings > Usage when one is on offer, usage credits billed at API rates once the included allowance is spent, and a Max plan. The routes that cost nothing work before the wall or after a model-specific one: switching the chat to Sonnet or Haiku, lowering effort, and starting a fresh chat; a session or weekly limit is shared across every model, so switching does not clear it. A second app that reaches Claude on a per-message meter covers the days the weekly cap is gone.
Why does Claude say the usage limit is reached when I have credits?
Because usage credits only take over once they are switched on. Anthropic bills them at standard API rates after your plan's included session or weekly allowance is spent, and the toggle lives in Settings > Usage. A limit reset is a different thing again: it refills the included allowance and does not touch your credit balance.
Does the Claude usage limit reset at midnight?
No. The session limit resets five hours after the first message of the session, at whatever time of day that falls, and Claude prints that time in the limit message. The weekly limit resets at a fixed time each week that Anthropic assigns to your account, which is shown in Settings > Usage.
