When Your Claude Max Limits Reset

When your Claude Max limits reset

Your Claude Max limits reset on two separate clocks, and only one of them follows the calendar. The session limit runs on a rolling window that your own activity opens, so it closes at a different time every day. The weekly limit resets at a fixed day and time that Anthropic assigns to your account, and that one never moves. If you keep running out at three in the afternoon, the thing to change is where the window sits, because you are the one who opened it.

Take the mechanism from this page and take the numbers from your own account. Anthropic publishes the shape of the limits. Your app and your terminal publish your position inside them. No article can tell you how much capacity you have left right now, and any that tries is guessing.

The session window is rolling and the weekly window is fixed

Anthropic's Max plan help page states the session cadence plainly. Checked September 2026:

Your session-based usage limit will reset every five hours.

The same page does not name an hour of the day for that window, because there is not one. The Claude Code documentation describes the equivalent seat allowance on Team and Enterprise plans as resetting on a rolling five-hour window and a weekly window. Rolling means the window is anchored to your usage. It does not begin at midnight, and it does not begin at the top of the hour.

The weekly limit is the other kind of clock:

The weekly limit resets at a fixed time each week that is assigned to your account. Your reset day and time stay the same regardless of when you start using Claude or when your subscription begins, and you receive your full weekly allowance each cycle.

So you decide where the session window sits, and you do not decide where the weekly reset sits. The two also fail differently. A session limit stops you for part of an afternoon. A weekly limit stops you in the middle of a week you already planned, and no change of habit gets it back before its assigned time. The general shape of these limits across every paid plan is set out in how Claude usage limits are structured.

Both limits count across models. Anthropic's Claude Code documentation says a session or weekly limit is shared across all models, so you cannot restore access by switching models with /model. A model-specific limit is a different thing. After You've hit your Opus limit, switching to a model outside that family with /model does keep you working.

Where the app and the CLI show your reset time

Several places report your reset time, and every one of them is better than a number you read in an article.

In the web and desktop app. Open Settings, then Usage. Anthropic's help page says you can see your next reset time in Settings > Usage. The Current session row shows how much of the session limit you have used so far, plus the amount of time remaining in the session. That remaining time is the countdown to act on. Watch it once and the rolling behaviour becomes obvious, because it does not land on a round local hour unless you happened to start on one.

In Claude Code, the CLI (command line interface). Run /usage. On a Pro, Max, Team or Enterprise plan it shows your plan usage bars alongside a breakdown of what has been consuming the allowance. Press d or w to switch that breakdown between the last 24 hours and the last 7 days.

One caveat explains a gap that confuses people. The breakdown figures are approximate and computed from local session history on that machine, so work you did on another laptop or in the claude.ai app is missing from them. The usage bars still reflect your real account. A bar that looks fuller than the breakdown can explain is usually your other machine.

When the usage request itself fails, most often because the usage endpoint is rate limited, /usage falls back to the last bars it loaded on that machine within the past 60 minutes and prints a Showing last-known usage note, with r to retry. Stale bars look exactly like fresh bars unless you read that note.

In the limit message itself. When you reach the ceiling, Claude Code prints You've hit your session limit or You've hit your weekly limit, and the message states when that window resets. On Claude Code v2.1.234 or later, an interactive session signed in with a claude.ai subscription then waits and continues the task on its own:

Usage limit reached · continuing automatically at 3:45pm · esc to cancel

That line is the most useful reset time you will get, because it is the moment Claude Code itself will start work again. At the reset it changes to Usage limit reset · continuing automatically and picks the task up where it stopped, using a fixed prompt rather than resending your last message. If your computer slept for more than about 30 minutes across the reset, the line reads Your usage limit has reset · press enter to continue instead, and nothing moves until you press Enter. Esc at an empty prompt cancels the wait, and /rate-limit-options offers it again.

Run claude --version if none of that appears. The automatic wait is a recent feature, so an older binary stops at the limit and leaves you to come back on your own.

Why a heavy session empties the window faster

The window measures tokens, not messages, so two afternoons with the same number of prompts can cost very different amounts. Claude Code sends your full conversation with every request, and each time Claude uses a tool it sends another request carrying that batch of tool results. A one-line question inside a session that has been open all day still draws usage for the whole conversation.

Anthropic's cost documentation names the causes you can act on:

• Long context. A large conversation is re-read on every turn, at the cached token rate.
• Cache misses. Your first message after a break longer than the cache lifetime reprocesses your full context. That lifetime is one hour on a subscription, and it drops to five minutes once you are drawing on usage credits.
• Scheduled tasks. A scheduled task fires on its interval even while the session sits idle, sending your full context each time.
• Agent teammates. Each active teammate runs its own context window and keeps consuming tokens until it exits.
• Compaction. /compact reads the conversation it summarises, so compacting a large context is itself a large request. /clear costs nothing.

You do not have to guess which one is hurting you. The /usage breakdown flags any behaviour that accounts for 10% or more of your recent usage, and it attributes usage to skills, subagents, plugins and individual MCP (model context protocol) servers. If cache misses are flagged, the session you left open over lunch is the cause: the first message after the break paid for the whole history again. The mechanism underneath all of it is covered in how Claude Code spends tokens on your codebase.

Planning a UTC+8 working day around the window

UTC+8, which is coordinated universal time plus eight hours, covers Beijing, Singapore, Hong Kong, Taipei, Kuala Lumpur and Perth. The session window does not care about your offset, and that is the point. It cares about when you started.

Here is the useful consequence. If your window opens with the first message of the morning at 09:00, the countdown in Settings > Usage tells you when it closes, and the next window runs from there. Your day now has a boundary you placed on purpose, instead of one that surprises you at 15:20 with a half-finished migration on screen.

Four habits follow from that.

1. Open the window with the real job. A throwaway question at 08:20 while the coffee is brewing anchors the boundary in the middle of your afternoon. The same question asked after the morning's heavy session costs you nothing.
2. Check the bars before you start something big. If Settings > Usage shows the current window mostly spent with an hour left, do not start a large refactor into it. Do the reading and the small fixes, then open the big job in the fresh window.
3. Split long jobs at the boundary deliberately. Before the window closes, ask for a written plan or a checkpoint file. The next window then starts from a short document instead of a long conversation, which is cheaper and easier to pick up. Run /clear before you hand that document back.
4. Let it wait when the boundary is close. Automatic continue parks the task and resumes it at the reset while you do something else. Claude Code ends a wait it started on its own if the reset moves more than 24 hours out, so in practice it covers a session boundary and not a weekly one.

The weekly limit needs a different habit, because you cannot move it. Find your assigned reset day and time in Settings > Usage and write it down somewhere you will see it. The day before your weekly refill is the tightest day of your week, every week, however carefully you place your session windows. Schedule the large migration for the day after the refill. If your assigned time falls in the middle of your night at UTC+8, that is good news, because you start most mornings full.

What the higher Max tier changes about hitting the wall

Anthropic sells two Max tiers and describes each one on its Max plan page relative to the Pro plan. The tier changes how much you can spend inside a window. It does not change the shape of the window or the day your weekly allowance refills.

That distinction decides whether an upgrade helps you. If you hit the session ceiling twice a week on genuinely heavy days, a larger allowance buys room inside each window and you will meet the wall less often. If you hit it at the same point every single day inside the same long-running session, the pattern is your context rather than your plan, and a larger allowance moves the same wall an hour later. Fix the session hygiene first, then measure a normal week with /usage before you pay more.

The gap between the tiers in daily use is compared in the Max 5x and Max 20x tiers side by side, and the questions worth answering first are collected in what to check before upgrading your Max plan. If you are billed in the region this timing advice is written for, the sticker price is not what leaves your account, and what Claude Max actually costs in Taiwan and Hong Kong works through the currency and tax side of the same decision.

You are already locked out. What now?

The reset time is in the message on screen, so read it before you do anything else. Then take one of two exits. Wait in place and let automatic continue resume the task at the reset, or turn on usage credits with /usage-credits and keep working past the plan limit at a metered cost. A model-specific message has a third exit, since switching model families with /model keeps you going. Each option has conditions attached, and they are worked through in what to do when Claude says you have hit your limit.

FAQ

Does my Claude Max session limit reset at midnight?

No. Anthropic describes the session limit as resetting every five hours on a rolling window, and that window is anchored to your usage rather than to the calendar. Midnight and the top of the hour have no effect on it. The weekly limit is the one with a fixed schedule, and Anthropic assigns that day and time per account. Open Settings > Usage to see yours.

Where do I find my exact reset time?

Settings > Usage on claude.ai shows your next reset time and the time remaining in the current session. In Claude Code, /usage shows your plan usage bars, and the limit message itself states when the window resets. On Claude Code v2.1.234 or later, the waiting line names the resume moment directly, for example continuing automatically at 3:45pm. Trust those over any figure printed in an article, this one included.

Why did my limit run out faster today than yesterday?

Usage is counted in tokens rather than in messages. A long conversation is re-sent with every request, and every tool call adds another request carrying its results, so an eight-hour session with fifty prompts can cost far more than a fresh session with the same fifty prompts. Run /usage and read the breakdown. It flags any behaviour, such as long context or cache misses, that accounts for 10% or more of your recent usage.

Does switching models get me back to work?

It depends on which message you got. Session and weekly limits are shared across all models, so /model will not restore access after You've hit your session limit. A model-specific message such as You've hit your Opus limit is different, and switching to a model outside that family keeps you working until the shared window becomes the problem instead.

Does the higher Max tier make my limits reset sooner?

No. The tier changes the size of the allowance inside the window. The five-hour session cadence and your assigned weekly reset time stay exactly the same. Upgrading means you reach the ceiling less often, and it does not clear a ceiling you have already reached. Measure a normal week with /usage first, so you know which of the two limits is actually stopping you.