Claude Code Pro plan usage limits, explained

Pro costs $20 a month ($17/mo billed annually), includes Claude Code plus access to Opus, Sonnet, Haiku, and Fable, and runs on two gates: a five-hour rolling window, and a weekly allotment that keeps Opus on its own dedicated ledger instead of pooling it with the rest. What Pro doesn't come with is a published size for either gate - Anthropic confirms how the mechanic works, not how big it is, and that gap is exactly where most of the confusion around this plan comes from.

What Pro actually includes

Pro is the entry paid tier: $20/month billed monthly, or $17/month ($200 upfront) on the annual plan. It bundles Claude Code with access to Claude on the web, desktop, and mobile, plus every current model tier - Opus, Sonnet, Haiku, and Fable - rather than gating specific models behind a higher plan. What Pro doesn't include is Max's bigger room inside those same two budgets, or the priority access and Cowork features Max adds on top.

That's the whole shape of it: one flat price, full model access, and a usage ceiling that's smaller than Max's but works exactly the same way underneath.

The two limits that gate a Pro session

Every Claude subscription plan - Pro, Max, Team, and seat-based Enterprise - runs on the same pair of gates. The first is a five-hour rolling window: it resets on a rolling basis, and Claude Code shows you the exact reset time the moment you hit it. The second sits on a weekly clock that keeps Opus in its own lane apart from every other model, so a week spent mostly in Sonnet leaves your Opus room untouched, and vice versa.

Both budgets are shared across surfaces, not per-app - usage on claude.ai, Claude Code, and Claude Desktop all draws from the same underlying pool, according to Claude's own help center. A long chat on the website this morning counts against the same weekly number Claude Code is about to hit tonight. What actually drains either budget: message length, attachment size, how long the conversation has run, which tools and connectors get called (Anthropic's docs specifically call these "token-intensive"), which model you're using, and the effort level you've selected.

What Anthropic confirms, and what it doesn't

I fetched Pro's pricing page and both of Claude's relevant help-center articles directly for this guide rather than working from what other posts claim about them, and the gap between what's confirmed and what isn't matters more than either fact alone. Anthropic documents that the two budgets exist, how they're structured, and what drives them faster. It does not publish an exact message count, token count, or size figure for either window, on any plan tier - Pro included.

That matters because a fair amount of what circulates about this topic states specific numbers with more confidence than the source material supports: a round message count, a claim that a cap size doubled on a specific date, a precise token ceiling. None of that is unreasonable to estimate, but it isn't documented, and treating an estimate as a confirmed spec is how "why did my number not match the article" threads get started. If you want the real shape of your own ceiling instead of an estimate, /usage inside Claude Code shows progress bars for both windows against your actual account - that number is the one that's real.

Pro next to Max, side by side

| | Pro | Max 5x | Max 20x | | --- | --- | --- | --- | | Price | $20/mo ($17/mo annual) | $100/mo | $200/mo | | Session + weekly room | Baseline | 5x Pro's room | 20x Pro's room | | Models included | Opus, Sonnet, Haiku, Fable | Same, larger room | Same, largest room | | Claude Code access | Included | Included | Included | | Fits | Light to regular daily use | Regular-to-heavy daily coding | Heavy parallel work, multiple subagents |

Max doesn't add anything to what a session can do - same models, same two-budget mechanic, same Claude Code. What it buys is more room inside both budgets before you hit either ceiling. Whether that room is worth the jump in price is a separate question from what Pro itself includes, and it's one worth checking your own setup against before answering.

Before you assume Pro is the problem

A session that runs out of room in minutes instead of hours is rarely Pro being genuinely too small - it's usually one specific habit burning the budget faster than the work justifies. The full breakdown of what actually drives that is worth reading if you're already hitting the wall regularly, but the short version is: check /usage first to see which budget you're actually close to, then look at whether cache_create is behaving normally, whether your CLAUDE.md and MCP servers are carrying dead weight, and whether that matches this project's own measured recovery number before you conclude Pro's ceiling is the real constraint.

When Pro genuinely isn't enough

Not every Pro lockout is a setup problem. A long migration, an all-day pairing session, or several subagents running in parallel spends a real budget honestly, and trimming a CLAUDE.md doesn't manufacture more room out of a genuinely full one. If /usage shows clean behavior - cache reads dropping after the first message, no dead MCP servers, a lean CLAUDE.md - and you're still hitting the wall on real work, that's Pro's actual ceiling, not waste, and Max 5x or Max 20x is the honest next step rather than another round of trimming. The cost calculator is a fast way to see what your own session pattern would cost across Pro, Max, and raw API rates before you commit to any of them.

FAQ

• How much does the Claude Pro plan cost? $20/month billed monthly, or $17/month ($200 billed upfront) on the annual plan, confirmed on claude.com/pricing.
• What models does Pro include? All of them at time of writing - Opus, Sonnet, Haiku, and Fable - rather than gating specific models behind Max. The difference between tiers is how much room you get inside the two usage budgets, not which models you can reach.
• What's the exact Pro plan usage limit, in messages or tokens? Anthropic doesn't publish one. Its own docs confirm the two-budget structure - a 5-hour rolling session window and a separate weekly cap - without stating a specific number for either. Treat any exact figure you see elsewhere as an estimate, and check /usage for your own account's real numbers.
• Does Pro include Claude Code, or is that a separate cost? Included. Pro is one flat subscription that covers Claude on the web, desktop, mobile, and Claude Code, with no separate metering for the terminal tool.
• Is Opus limited differently than other models on Pro? Yes. Opus runs on its own dedicated weekly ledger, apart from Sonnet, Haiku, and Fable - heavy use of one doesn't draw down the other.
• How is Max different from Pro if the models are the same? Max doesn't change what a session can do - it changes how much room you get inside the same 5-hour and weekly budgets before hitting the ceiling. Max 5x is roughly 5x Pro's room for $100/mo; Max 20x is roughly 20x for $200/mo.

Anthropic's docs can tell you the two gates exist; they can't tell you which one your own sessions are burning, or why. The free scan looks at that part - the dead MCP servers, the oversized CLAUDE.md, the exploration nobody offloaded - and hands back a plain answer to whether cleanup buys you more room than upgrading would.

Made by Neo

UsageCut by ClockedCode - not affiliated with Anthropic. The figures on this page are measured on real Claude Code sessions and labeled measured or estimated where it matters.