Claude usage limits: the 5-hour session limit, the weekly limit, and when they reset

Claude's usage limits are two windows. A session limit resets every five hours, and paid plans add a weekly limit that resets at a fixed time each week. Claude Code, claude.ai and Claude Desktop draw on the same limits, and Anthropic publishes no message or token count for any plan, only how the plans compare.

What is the Claude 5-hour limit?

It is the session limit. "Every plan has usage limits that reset on a rolling five-hour session window," Anthropic's pricing page says. What a message costs against it depends on the length and complexity of the conversation, the model and the features in use, "so there's no fixed message count".

Anthropic sizes each plan against another, in the table at the top of this page: Pro gives "at least 5x more usage per 5-hour session than Free", Max 5× and Max 20× five and twenty times Pro, a Team standard seat more than Pro and a premium seat five times a standard one (which seat a developer needs: Claude Team plan).

The Claude weekly limit

On top of the session limit, Pro, Max and Team seats have a weekly limit that applies across all models. Max plans also have a separate weekly limit for Fable. Claude Code can warn you before a window runs out (You've used 85% of your session limit).

Every five-hour window also draws on the week. Anthropic's Claude Code docs put it plainly: "A single burst of heavy activity, such as a large workflow fanout, can exhaust the weekly allowance before the session window resets." Once the week is used up, a fresh five-hour window does not help.

When does the Claude Code usage limit reset?

• The session limit resets five hours into its window.
• The weekly limit resets "at a fixed time each week that is assigned to your account" — not seven days after you start.
• To see the time: the limit message carries it (You've hit your weekly limit · resets Mon 12:00am), /usage in Claude Code shows it beside each bar, and so does Settings > Usage on claude.ai. The status line receives it as resets_at.

Anthropic also gives eligible plans an occasional limit reset: under Settings > Usage on the web or in Claude Desktop, Reset for free sets your session or weekly limit back to full. The button is not in Claude Code, but the limits are shared, so the reset applies there too. An unused reset expires on the date shown in the offer.

Session, weekly and model limits in Claude Code

Claude Code names the limit you hit:

You've hit your session limit · resets 3:45pm
You've hit your weekly limit · resets Mon 12:00am
You've hit your Opus limit · resets 3:45pm
You've hit your Sonnet limit · resets 3:45pm

The session and weekly limits are shared across all models, so switching models does not help. The Opus and Sonnet limits apply only to that family: /model to a model outside it keeps you working, though the next request re-reads the whole conversation without the cache.

How many messages per day does Claude allow?

Anthropic documents no daily limit and no count of messages or tokens, per day or per five hours. Use is measured in the two windows, and Anthropic keeps the right to limit it "in other ways, such as weekly and monthly caps or model and feature usage". The nearest thing to a number is your own: the percentage of each window that /usage shows.

Claude free plan usage limits

The Free plan has the same five-hour session window; the weekly limits are added on paid plans. "Free covers everyday questions," in Anthropic's words, and Pro gives at least five times Free's usage per session. Claude Code comes with the paid plans.

Claude peak hours

Anthropic's help articles on usage limits do not say how limits change at busy times of day. What its plan table does say: Max, Team and Enterprise include "Priority access at high traffic times"; Free and Pro do not.

Claude Code and claude.ai share one limit

Anthropic counts every surface against the same allowance: claude.ai, Claude Code and Claude Desktop. A long research chat in the browser leaves less for code in the terminal, and the other way round.

What happens when you hit the limit

• Wait. In Claude Code 2.1.234 and later, an interactive session signed in with a subscription waits and continues the task on its own when the reset is less than 24 hours away: Continuing automatically at 3:45pm · esc to cancel. For a weekly reset days out, pick Wait here, then continue automatically from /rate-limit-options.
• Wrap-up. If the session limit arrives partway through a response, Claude Code may finish the step in progress and show Usage limit reached · wrapping up — up to once a week on Pro, each time on Max. Anthropic is rolling this out gradually.
• Usage credits, with /usage-credits, billed at standard API rates; your work carries on instead of stopping.
• Upgrade: Pro to Max 5×, Max 5× to Max 20×.

With ANTHROPIC_API_KEY set, Claude Code bills that key instead of your plan once you approve it, and always in -p mode. Usage is then billed per token instead of counting toward your plan. /status shows which one is active.

How close I came, measured

My own machine, from Caprock's record of the windows Claude Code reported on the days its status line was installed. The highest five-hour window was 17% on Max 20× ($200) and 47% on Max 5× ($100); the highest week 27% on Max 20×, and 24% on Max 5× after four days.

On Max 20× I was never near either cap. On Max 5× my work ran a five-hour window to 47% and used a quarter of the week in four days. Those are 9 days with readings — 19 August and 6–9 September on Max 20×, 29 September to 2 October on Max 5× — not a full record. From 20 August to the end of September, every one of the 29 turns Claude Code ended on an error was something other than a usage limit.

See how close you are

• /usage in Claude Code shows your plan's usage bars, when each resets, and what has been using them over the last day or week.
• Settings > Usage on claude.ai shows the same limits for the whole account.
• The status line gets rate_limits.five_hour and rate_limits.seven_day — the used percentage and reset time — on Pro and Max, after the first response. The status line guide shows how to put them on screen.

Caprock keeps those readings, for every session on the machine whether or not it started it. With its status line installed (caprock statusline install), the Cost screen shows how much of each window is used and when it resets, and a forecast appears only when your measured pace would hit the limit before the window resets. Every dollar on that screen is a list-price estimate, as Claude Code's own /usage is, next to the plan's dollars and the share of the weekly limit the work used. Codex's windows sit below Claude's in the same panel, from its own session files.

Codex limits, for comparison

OpenAI publishes estimates where Anthropic publishes ratios: on ChatGPT Plus and Business, a number of local messages per five hours for each model, which it says "are not fixed message limits". Codex usage limits covers Codex's windows by plan; Codex vs Claude Code compares the two plans in full.

Install Caprock — free, open source, nothing leaves your machine; it keeps your windows from the moment its status line is installed with caprock statusline install, which the headless caprock up also offers.