How the weekly limit actually works
Every paid Claude plan has two meters. The 5-hour window starts with your first message and caps how much you can do in that window (roughly 10–40 Claude Code prompts on Pro, 50–200 on Max 5x, 200–800 on Max 20x). The weekly limit caps total usage across all models over 7 days, with a separate, smaller Opus allowance on Max plans. Both are measured in tokens, not prompts or hours — Anthropic publishes hour ranges because that's what people can plan with, and the width of each range is entirely about context size.
That's the key to this calculator. A session you /clear between tasks sends a few thousand tokens per request and lands you at the top of the range (80 Sonnet hours on Pro). A session that's been open since morning sends 100,000+ tokens of history with every single prompt and lands you at the bottom (40 hours). Same plan, same hours at the keyboard, double the difference.
Published ranges
| Plan | Prompts / 5 h | Sonnet hrs / week | Opus hrs / week |
| Pro ($20) | ~10–40 | ~40–80 | — (no Opus in Claude Code) |
| Max 5x ($100) | ~50–200 | ~140–280 | ~15–35 |
| Max 20x ($200) | ~200–800 | ~240–480 | ~24–40 |
What burns the limit fastest
- Long context. The single biggest factor. Claude Code re-reads your whole conversation every turn (at the cached rate, but it still counts). Clear between unrelated tasks; use
/compactwith instructions on what to keep. - Opus. About 2.5× the draw of Sonnet per hour, and it has its own weekly cap. Max 5x auto-switches to Sonnet at 20% of usage, Max 20x at 50%.
- Agent teams and parallel sessions. Each teammate has its own context window; Anthropic notes agent teams use roughly 7× the tokens of a normal session when teammates run in plan mode.
- Cache misses. Come back after more than an hour and the first message reprocesses the whole context at full price. Subscribers get a 1-hour cache lifetime; usage credits drop it to 5 minutes unless you set the TTL yourself.
- Scheduled tasks and loops. A task that fires every 10 minutes sends your full context every 10 minutes, even while you're at lunch.
- Extended thinking. Thinking tokens bill as output; lower the effort level for simple tasks.
- Chat, Cowork and Design. Same pool. A heavy Cowork afternoon can take a third of a Pro week.
When you do hit it
Claude Code tells you when the window resets and, on recent versions, can wait and continue the task automatically when it does. Or turn on usage credits (/usage-credits) and keep working at API list rates — Sonnet roughly $1.50–8 per active hour depending on context, Opus about 2.5× that — with a monthly spend limit you set and a hard $2,000/day cap. If you're hitting the weekly limit every week on Pro, Max 5x costs less than the credits would. If you're hitting it on Max 5x, check the context habits above before paying for 20x; most people find /clear buys back more than the upgrade.
Frequently asked questions
What's the Claude Code weekly limit?
Pro ~40–80 Sonnet hrs; Max 5x ~140–280 Sonnet + 15–35 Opus; Max 20x ~240–480 Sonnet + 24–40 Opus. Measured in tokens, so context size sets where you land.
When does it reset?
Every 7 days from the start of your usage window; the exact time is in /usage and in the limit message. The 5-hour window is rolling.
Why did I hit it so fast?
Long sessions with big contexts, Opus, parallel agents, or chat usage on the same account.
Does switching models help?
If you hit the Opus-specific cap, switching to Sonnet keeps you working. If you hit the all-model weekly cap, no.
How much do usage credits cost?
API list rates: roughly $1.50–8 per hour on Sonnet, 2.5× on Opus, up to $2,000/day.
Does the limit apply to the API?
No — API keys have rate limits (tokens per minute) and a bill, not a weekly cap.