Claude Usage Limits: Pro & Max Caps Explained (Oct 2026)

Claude usage limits work on two clocks: a 5-hour session window and, on paid plans, a weekly cap. When either runs out, Claude stops until it resets, unless you upgrade, use a limit reset, or turn on usage credits billed at API rates. Anthropic explains the rules in its Help Center article on how usage and length limits work and on the Claude pricing page.

This page is kept at one URL and updated as Anthropic changes things. It covers every plan, what counts against your limit, the Claude Max limits A/B test reported this week, and how to stretch what you have. It was written by a Sumus AI Employee.

How Claude usage limits work

• 5-hour session limit. Every plan, including Free, has a usage allowance that resets on a rolling five-hour window.
• Weekly limit. Pro and Max add a weekly limit across all models. It resets at a fixed day and time assigned to your account, no matter when you subscribed; check it in Settings > Usage.
• One shared pool. Usage on claude.ai, Claude Desktop, mobile and Claude Code counts toward the same limit.
• No fixed message count. How far you get depends on conversation length, attached files, the model, effort level and features such as tools and connectors.
• Extra caps at Anthropic's discretion. Anthropic says it may limit usage in other ways, "such as weekly and monthly caps or model and feature usage," to manage capacity.

Claude usage limits are separate from length limits. A length limit is the context window of a single chat (up to 1M tokens on the newest models on paid plans). Hitting it means starting a new chat, not waiting.

Claude usage limits by plan

One detail about Claude usage limits trips people up: the 5x and 20x multipliers apply to the per-session allowance. Anthropic's Help Center describes Max 20x as "20 times the Pro plan's per-session usage allowance," while the weekly cap is set separately. That is why a recurring question on X and Reddit is whether the $200 plan really delivers 20x Pro's weekly usage.

Claude Max limits: the October 2026 A/B test

On October 22, 2026, Claude Max subscribers flooded X with reports that their meters were draining faster. Claude Code engineer Lydia Hallie replied to one user: "our limits haven't changed," and asked for a /feedback ID. About 70 minutes later SemiAnalysis posted that it had detected an A/B test on usage limits affecting some accounts.

• Both accounts ran Opus 5.5 on the same day on Max 20x.
• Account B's 5-hour window held 2,076 million agentic tokens, 19% more than Account A's 1,740 million.
• Account B's weekly budget was 5,266 million tokens versus 6,579 million, 20% less.
• Because the weekly cap binds heavy users over a month, SemiAnalysis values Account B at about $9,385 of API-equivalent usage per month versus $11,726, roughly 47x versus 59x the $200 price.
• SemiAnalysis says it saw a similar test earlier in October, which a provider described as an "extremely tiny" A/B test on "how to better balance when people hit limits."

In follow-up replies, Hallie asked users whether they often return to long sessions after the prompt cache expires (1 hour+), run many subagents, and which model and effort level they use, all of which burn more usage. The practical takeaway: if your Max plan feels thinner this week, watch the weekly meter, not the five-hour one.

Recent changes to Claude limits

• Sep 4, 2026: Anthropic reset weekly limits for all Max subscribers after the Fable 5.1 launch.
• Mid-September 2026: users reported a +50% weekly promo on Max ending and being replaced by a permanent +25%.
• October 2026: one-time limit resets appeared in Settings > Usage, with expiry dates; some users reported banked resets expiring Oct 22.
• Oct 22, 2026: the Max A/B test reported by SemiAnalysis.

What to do when you hit Claude usage limits

• Wait for the reset. Claude usage limits refresh in two steps: the session limit refreshes five hours after it starts; the weekly one at your assigned time.
• Use a limit reset. If you have one, go to Settings > Usage on the web or Desktop and click "Reset for free." It restores either the session or weekly limit, can't be undone, and isn't available in mobile or the Claude Code terminal.
• Turn on usage credits. Pro and Max users can keep going at standard API rates, with monthly spend caps, auto-reload and a $2,000 daily redemption limit.
• Upgrade. Pro to Max 5x or 20x is prorated for the rest of your billing cycle.

How to make Claude usage limits last longer

• Start a new chat for new topics; long threads and automatic context management cost more.
• Pick a lighter model or lower effort level for routine work, and turn off extended thinking when you don't need it.
• Disconnect tools and connectors you aren't using; they are token-intensive.
• Keep project instructions short and remove unused project files.
• In Claude Code, run /usage to see where your usage goes, and avoid resuming long sessions after the cache has expired.

Where Sumus fits

Yes, a Sumus AI Employee can do this kind of research for you. This guide is an example: it read Anthropic's Help Center and pricing pages, pulled the SemiAnalysis numbers from X, built the tables and published the page. Sumus AI Employees run on models like Claude Opus 5.5, Fable 5.1 and GPT-6 in your team's group chat, take tasks in plain words, bring in other agents for subtasks and hand back work everyone can review.

Give one a standing job such as "check Anthropic's limits pages every week and tell us what changed." See how it works on the Sumus AI Employee page.

What we're watching

• Whether Anthropic confirms or ends the Max weekly-limit A/B test on Claude usage limits.
• Any change to the Pro and Max multipliers or the Fable share of weekly limits.
• New limit resets or promotions announced on Anthropic's official channels.