Home · Guides · Claude Code usage limits
Claude Code usage limits, explained
The rolling 5-hour window, the weekly cap, the separate Opus limit — how they actually work, what the limit messages mean, and when things reset.
Checked against official docs: July 13, 2026The two limits (plus one for Opus)
Every Claude subscription — Free, Pro, Max — has two independent usage limits, and they are shared across claude.ai, Claude Code and the desktop app. One usage budget, all surfaces:
- 5-hour window — rolling. It starts with your first message, not at midnight and not on a fixed schedule. Five hours after that first message, the window resets. official
- Weekly cap — a second ceiling on top, introduced in August 2025. It resets every 7 days at a fixed time for your account (shown in Settings → Usage). official
- Opus weekly limit — on Max plans, Opus has its own weekly budget with its own reset timer, separate from the all-models limit. Max plans also track a Sonnet-only weekly limit separately. official
Terminology note: these subscription usage limits are not the same thing as API rate limits (requests-per-minute for pay-as-you-go API keys). Anthropic's error reference explicitly separates the two — if you're being throttled on an API key, that's a different page of the docs.
Limits by plan
Anthropic deliberately doesn't publish fixed hour counts as a guarantee — capacity varies with message length, model, and codebase size. The most concrete official numbers ever given were the launch estimates from August 2025, and limits have only been raised since:
| Plan | Price | Weekly Sonnet estimate* | Weekly Opus estimate* | Opus access |
|---|---|---|---|---|
| Free | $0 | ~15–40 messages per 5-hour window community | — | No |
| Pro | $20/mo | 40–80 h | — | No |
| Max 5x | $100/mo | 140–280 h | 15–35 h | Yes |
| Max 20x | $200/mo | 240–480 h | 24–40 h | Yes |
* August 2025 launch estimates from Anthropic's announcement — treat as a floor, not a promise. See the change log below: 5-hour limits have been doubled and weekly limits boosted since.
2026 limit changes — the log
Anthropic moves these numbers regularly, usually upward, sometimes as time-boxed promotions. This log is why this page has a "checked" date:
- Jul 13, 2026 — the +50% weekly-limit promotion for Claude Code was extended to July 19, 23:59 PT. Applies automatically to Pro, Max, Team and seat-based Enterprise; Claude Code only (not claude.ai chat).
- May 13, 2026 — weekly Claude Code limits raised 50% as a promotion (originally through July 13).
- May 6, 2026 — 5-hour limits doubled for Pro/Max/Team, peak-hours reductions removed, Opus API rate limits raised, on the back of Anthropic's SpaceX compute deal.
- Aug 28, 2025 — weekly caps introduced (the launch estimates in the table above). Anthropic said fewer than 5% of subscribers would notice.
What happens when you hit a limit
Claude Code tells you exactly which ceiling you hit — these are the real messages from the official error reference:
You've hit your session limit · resets 3:45pm— the 5-hour window. Applies to all models; switching models won't help. Wait for the reset.You've hit your weekly limit · resets Mon 12:00am— the weekly cap. Also model-independent.You've hit your Opus limit · resets 3:45pm— only Opus is blocked. Run/modeland switch to Sonnet to keep working.
Since 2026 you can also buy extra usage instead of waiting: /usage-credits enables opt-in usage credits on Pro and Max plans, billed at standard API rates (with a configurable monthly cap and a $2,000/day ceiling).
"Why did my limit reset earlier / later than I expected?"
- The 5-hour window is anchored to your first message, so the reset time moves around with your work pattern. Two coding sessions a day = two different reset times.
- The weekly reset is fixed per account — check Settings → Usage for your exact day and time. Community measurements have sometimes observed shorter effective cycles (~72h) during promo periods, so if your weekly reset came "early", it was in your favor.
- There are real bugs, too: users have reported Opus budgets draining at 16% per half-hour (GitHub issue #8449) and abnormal drain waves in March 2026. If your numbers look impossible, it's worth checking the claude-code issue tracker before assuming you burned it all.
How to check your usage right now
- In Claude Code: run
/usage— 5-hour and weekly percentages with reset countdowns (in older builds this lived under/status). - On the web: claude.ai → Settings → Usage — progress bars for the 5-hour window, the weekly cap, and the separate Opus reset.
- Desktop app: the usage ring next to the model picker.
- Without asking: keep it in your menu bar, permanently — that's the whole reason AI Status Bar exists.
FAQ
Do Claude Code and claude.ai share the same limit?
Yes. Your plan has one usage budget shared across claude.ai chat, Claude Code and the desktop app. A heavy chat session eats into your coding window and vice versa.
Does switching models reset or avoid the limit?
The 5-hour and weekly limits count all models, so no. The one exception is the separate Opus limit: when Opus is exhausted, Sonnet still works until you hit the general ceilings.
Are the current limits permanent?
No — treat them as a moving target. The weekly boost running right now is a dated promotion (through July 19, 2026), and Anthropic has changed the baseline three times in 2026 alone. That's why this page carries a "checked" date and a change log.
Can I pay to keep going after hitting the limit?
Yes, on Pro and Max: run /usage-credits to enable usage credits, billed at standard API rates. On Team/Enterprise, extra usage goes through your admin.
How do I know when my limit resets without opening anything?
That's the menu-bar use case: AI Status Bar shows the exact reset time next to each window's percentage, and can notify you at 80% and 90% before you hit the wall mid-task.