Independent tracker for Claude usage-limit resets — time since the last one, a shared statistical estimate for the next, and full history with source links. Free JSON API for your own bots and dashboards.
Last reset 65 days ago (Max users only) · 2026-09-04
An estimate from past intervals, not a schedule.
Since last reset: 64.5days · Typical gap 8 days
Already past the typical gap (reference date 2026-09-12)
Estimate from whenreset.app's shared model. See the method → · how this is calculated
An estimate from past intervals, not a prediction — Anthropic has never committed to a reset cadence.
12 reset days on record (1 partial)
“We've just reset weekly limits for everyone on a Claude Max plan. With Fable 5.1 out and a long weekend ahead for many of you, we wanted to keep you building. I'd love to see what you build!”
Last reset for all users: 2026-09-01 (68 days ago)
Last banked reset credited: 2026-09-22 — a banked credit is redeemed by the account holder, not a reset Anthropic clears for everyone at once.
Context only — these don't change the estimate above.
One column per week, one cell per day.
Latest announcements, newest first. Times in UTC.
Opus 5.5 performs at the level of Fable 5.1, ~30% faster and ~40% cheaper than Opus 5 per task. In Claude Code: 5-hour session limits increase 20% today; Pro, Max, and Team users get a reset to use anytime (available in Settings → Usage, apply any time until Oct 22).
We've just reset weekly limits for everyone on a Claude Max plan. With Fable 5.1 out and a long weekend ahead for many of you, we wanted to keep you building. I'd love to see what you build!
With Fable 5.1 out today, we've also reset 5-hour and weekly limits for all users.
We've reset 5-hour and weekly rate limits for all users.
We've reset 5-hour and weekly rate limits for all users.
Now that Fable 5 is ready to build (again), we've reset everyone's 5-hour and weekly rate limits.
Sponsored · synced from howtok.net
NovaAPI describes itself as a global AI token relay platform offering unified access to OpenAI, Claude, Gemini, and Chinese LLMs through low-latency, always-on routing. It advertises transparent pricing with API calls discounted up to 15% off list rates, plus a referral program with ongoing rewards.
CodeKey describes itself as an AI API relay for developers, individuals, and teams, with access to Claude, GPT, DeepSeek, GLM, Kimi, Grok, and Gemini. It provides an English-language site, accepts WeChat Pay, Alipay, and USDT, and states that ¥1 paid gives $1 in platform credit. The operator also states that ordinary invoices are available from ¥200 without a fee and that refunds are available at any time without a fee.
These are sponsored links, synced from howtok.net's pinned relay-station listings — third-party API resellers, not Anthropic. Claude Reset has no affiliation with them; use your own judgment.
No account, nothing sent anywhere — this checks the free JSON API from your own browser.
Works only while this tab stays open. For background alerts, subscribe via Atom or calendar.
Use any RSS/Atom reader: Atom feed · add calendar.ics to your calendar app for past resets and banked-expiry reminders. Updates only when an official reset is announced — no other notifications.
Check the answer line at the top of this page — it looks at whether any reset was announced today (UTC). If today's date isn't shown as the last reset, nothing has been announced yet today.
Anthropic has never published a schedule for resetting Claude usage limits, so there is no date to point to. The most recent reset on record was September 4, 2026 (scope: Max users). Every reset recorded on this page was announced ad hoc. The percentages at the top come from whenreset.app's shared model and are an estimate from past announcements, not a schedule. Your own 5-hour and weekly windows roll over separately — see /usage-limits/ for when.
Check the answer line and the last-reset card at the top of this page. They show whether Anthropic announced a reset today (UTC), who it covered — everyone, or one plan only — and how long ago it was. Each entry in the timeline links to the original post, so you can read the announcement yourself.
Claude Code and claude.ai share one usage limit, so a reset applies to both. Nothing in Anthropic's docs lets you reset your limits on demand: your 5-hour session window and weekly window roll over on their own, and /usage in Claude Code shows when. The one thing you can trigger yourself is a limit reset, if Anthropic has credited one to your account — it appears in Settings → Usage on the web or in Claude Desktop (the “Reset for free” button isn't available in Claude Code itself). See /banked-reset/ and /usage-limits/.
A public announcement from Anthropic's official @ClaudeDevs account or a member of the Claude team that usage limits were cleared or credited for some or all users — not your own personal 5-hour/weekly window rolling over on its own schedule. Each row links to the original post.
The percentages come from whenreset.app's shared climatology model, used across our sister sites so readers see one consistent number — not something this site calculates on its own. It's the share of past reset intervals that fall within a given window; it doesn't change just because more time has passed since the last reset. See /forecast-method/ for the model and this site's own retired research formula.
No. Every reset on this page was announced ad hoc, usually tied to a model release, an incident fix, or as a one-off gesture. There is no fixed cadence to rely on.
Instead of clearing your usage immediately, Anthropic sometimes credits a reset you can apply yourself, whenever you choose, before an expiry date. See /banked-reset/ for what's known about how it works.
No — those roll on their own per-account schedule regardless of anything on this page. This site only tracks the separate, occasional global/partial resets Anthropic announces publicly. See /usage-limits/.
Public posts from Anthropic's @ClaudeDevs account and Claude team members, collected by the upstream howtok.net project. This site adds its own live signals: status.claude.com and GitHub issues on anthropics/claude-code.
Possibly — detection is keyword-based, and we'd rather miss one than misclassify something else. If you spot an announcement that isn't listed, it's an upstream howtok.net data question, not something this site edits by hand.