Claude Code Usage Limits: Pro vs Max, and How to Stay Ahead of Them
Claude Code doesn't charge you per prompt — it meters you. Your plan (Pro, Max 5x, or Max 20x) sets how much coding you can do inside two overlapping windows: a rolling 5-hour session and a longer weekly cap. Anthropic tunes these limits often, so the durable skill isn't memorising a number — it's understanding how the windows work and watching your own burn rate so a 'you've hit your limit' message never catches you mid-task.
Last updated July 11, 2026 · Limits change often — always check Anthropic's current numbers.
How Claude Code limits actually work
Claude Code doesn't bill per prompt — it meters your usage. Two independent counters run at the same time, and you're throttled the moment you hit either one: a rolling 5-hour session window and a set of weekly limits. Understanding both is the whole game.
The 5-hour session window
This is the short window. It's a rolling 5 hours that starts with your first prompt, not a fixed clock. Send your first message at 10am and that window resets at 3pm — however many messages you sent in between. It's the limit most people bump into during an intense afternoon of coding.
The weekly windows
Anthropic added weekly limits in 2025 (announced in late July, effective the end of August) to sit on top of the 5-hour window. There are actually two of them: one overall weekly cap that counts usage across all models, plus a tighter weekly cap for the most capable Opus-class models, which primarily affects the Max plans (they carry the most Opus access). The two can reset on different schedules.
Pro vs Max 5x vs Max 20x
Your plan is really a capacity multiplier. The paid tiers scale the same two windows up — Max plans simply give you a lot more room in each, plus the separate Opus-class weekly cap. The figures below reflect Anthropic's plans as of mid-2026 and change often, so treat them as the shape, not the letter of the law — check Anthropic's official pricing page for the current numbers.
| Plan (mid-2026) | Relative throughput | Weekly caps | Best for |
|---|---|---|---|
| Pro (~$20/mo) | Baseline Claude Code access | One overall weekly cap | Individual devs, lighter daily use |
| Max 5x (~$100/mo) | Roughly 5x Pro's usage | Overall + Opus-class weekly cap | Heavy solo developers |
| Max 20x (~$200/mo) | Roughly 20x Pro's usage | Larger overall + Opus-class weekly cap | All-day, multi-agent workflows |
What actually counts against your limit
It's not a simple prompt count. Two people who send the same number of messages can burn wildly different amounts, because usage depends on:
- ✦Which model you run — Opus-class models consume far faster than Sonnet
- ✦How long your conversation and context have grown
- ✦The thinking / effort level you've selected
- ✦Tool calls, file reads, and how much the agent explores
- ✦Large pasted files or whole codebases
- ✦How many agents you run in parallel
What happens when you hit a limit
- Claude Code stops and tells you which limit you've reached and when it resets.
- The session limit clears 5 hours after that window's first prompt — often a short wait.
- A weekly limit clears on its 7-day cycle; the overall and Opus-class caps can reset separately.
- Open Settings > Usage in Claude for the authoritative countdown — that's the only meter tied to your real quota.
- While you wait, you can switch to a lighter model (if only the Opus-class weekly cap is exhausted), upgrade your plan, or move to pay-as-you-go API / Console billing, which is separate from subscription limits.
Key takeaway: Anthropic's limits move — they added weekly caps in 2025 and permanently doubled the 5-hour window in May 2026. Don't memorise a number. Learn the two-window model, check Settings > Usage for the official meter, and watch your own burn rate so a limit never surprises you mid-task.
Stay ahead of the limit: watch your burn rate
The habit that keeps limits from ever being a surprise is simple: know your own burn rate. If you can see a heavy week building — or notice which model is eating your quota — you can pace the work instead of slamming into a wall at 4pm on a Thursday.
- ✦Know your heaviest hours and days so you don't schedule a big run right before a reset
- ✦Spot a ramp-up early, while there's still time to slow down or switch models
- ✦Split Opus-class vs Sonnet usage so you know which weekly cap you're pushing
- ✦Spread heavy agent runs across the week instead of one marathon session
How Token Forest shows your usage
Token Forest is a free desktop pet with an offline dashboard. It reads only your local Claude Code and Codex logs — token counts, model names and timestamps, never your source code or prompts — and works fully offline with no account. It won't replace Anthropic's official meter, but it gives you the part Anthropic doesn't: the trend and the pattern.
- ✦Burn rate — how fast you're spending across the day and week
- ✦Per-model breakdown — Opus vs Sonnet vs Codex, so you can see which cap you're pushing
- ✦A 26-week heatmap — your heavy days at a glance, so you can pace the week
- ✦Per-project and daily / weekly / monthly charts — where the usage is actually going
One honest caveat: for the exact 'how close am I right now,' Anthropic's Settings > Usage is authoritative — it's the only place tied to your real quota. Token Forest complements it with a private, local view of your trend so you can plan ahead. Its dollar figures are an offline estimate from a bundled price table, so they may differ from your real Anthropic or OpenAI bill.
Common questions
How do I know if I'm hitting my Claude Code limit?
Claude Code shows a message telling you which limit you reached and when it resets, and Settings > Usage in Claude gives the authoritative countdown. To see it coming, watch your burn rate — Token Forest's dashboard reads your local logs and shows a heavy week building before you actually hit the wall.
When does my Claude Code limit reset?
The 5-hour session window resets 5 hours after that window's first prompt — it's rolling, not a fixed clock. Weekly limits reset on a 7-day cycle, and the overall and Opus-class caps can reset on different schedules. The exact times are shown in Settings > Usage.
What's the difference between Pro and Max usage limits?
Max 5x gives roughly 5x and Max 20x roughly 20x more usage than Pro, and there's a separate, tighter weekly cap for the most capable Opus-class models — most relevant to Max subscribers, who have the most Opus access. Anthropic adjusts these numbers periodically, so check the current limits on your account.
Does hitting my weekly limit stop me completely?
It stops the usage that's capped. If only the Opus-class weekly cap is exhausted you can often keep working on a lighter model like Sonnet. If the overall weekly cap is hit, you wait for the reset, upgrade your plan, or switch to pay-as-you-go API / Console billing, which is metered separately from your subscription.
Can Token Forest tell me exactly how much of my limit is left?
No — Anthropic's Settings > Usage is the only authoritative meter for your remaining quota. Token Forest reads your local token logs to show burn rate, a per-model split, and a heatmap: the trend picture that helps you plan and pace, alongside Anthropic's official number.
See your own burn rate
Token Forest turns your local Claude Code and Codex logs into a private usage dashboard — burn rate, per-model split, and a 26-week heatmap — so a limit never catches you mid-task. Free, offline, no account.