Aller au contenu
Discord UsageMeter : 1 mois de Pro offert

Claude usage limits explained: 5-hour, weekly and Fable

Claude has a rolling 5-hour session limit (none on a verified Ultra account) and a weekly limit shared by chat, Desktop and Claude Code. How each resets.

Claude limits your use in two layers: a 5-hour session limit that rolls from your first message, on Free, Pro and Max, and a weekly limit on every plan, which on paid plans resets at a fixed time each week. Ultra, the former Max 20x, has no 5-hour limit once Anthropic has verified the account (What is the Ultra plan?). Both limits are shared by claude.ai, Claude Desktop, the mobile apps and Claude Code, and both measure how much Claude processes rather than how many messages you send: Anthropic says there is "no fixed message count" (Claude pricing FAQ).

How the 5-hour session limit works

Anthropic describes it as a "rolling five-hour session window" and applies it to Free, Pro and Max (Claude pricing FAQ). In practice the window opens with the first message you send while no window is running and closes about five hours later, whatever you do in between: send your first message at 09:00 and the session resets at 14:00. Then nothing runs until your next message opens a new window. Anthropic doesn't document the exact start rule, so trust the reset time Claude shows you.

The 5-hour window starts with your first message, not at a fixed time

Example: first message at 09:00 → reset at 14:00. The next message, at 15:00, opens a new window.

Example
Two 5-hour windows on one dayFirst message at 09:00: window 1 runs until 14:00. Nothing runs between 14:00 and 15:00. The 15:00 message opens window 2, which ends at 20:00.WINDOW 1 · 09:00 → 14:00WINDOW 2 · 15:00 → 20:00↺ reset 14:00nothing runs until your next message1st message → the windowstarts08:0010:0012:0014:0016:0018:0020:0022:00dot size ≈ message sizeTwo 5-hour windows on one dayFirst message at 09:00: window 1 runs until 14:00. Nothing runs between 14:00 and 15:00. The 15:00 message opens window 2, which ends at 20:00.08:0010:0012:0014:0016:0018:0020:0022:00WINDOW 109:00 → 14:00↺ reset 14:00nothing runs until yournext messageWINDOW 215:00 → 20:00
Session windows in this example
PeriodFromTo
Window 109:0014:00
No window14:0015:00
Window 215:0020:00
Figure 1. Example times. A session lasts about five hours from the message that opened it; Settings > Usage shows the exact reset time.

Two things follow from that:

  • Pacing matters more than volume. A burst of long messages early in the window can fill it in an hour and leave you waiting for the rest.
  • The time left matters as much as the percentage. For example, 80% used with ten minutes left is fine; 80% with four hours left means slowing down or changing model.

On paid plans, Settings > Usage shows the session as Current session, with the time remaining (Usage limit best practices). The Free plan runs on the same five-hour logic, but "the number of messages you can send will vary based on demand" (Get started with Claude).

Ultra is the exception: "No 5-hour limit for verified accounts", says the pricing page (Claude pricing). Anthropic verifies accounts automatically from signals such as account and payment history (verifying a phone number at checkout doesn't remove the limit); until then the account keeps a temporary five-hour limit. Once verified, Settings > Usage shows no Current session bar (What is the Ultra plan?).

How the weekly limit resets

Every plan has a weekly limit that "applies across all models"; the pricing page lists Free with "Limited usage every 5 hours and every week", without a figure (Claude pricing). On paid plans it is not a rolling seven days: it resets at a fixed time each week assigned to your account, and that day and time "stay the same regardless of when you start using Claude or when your subscription begins" (What is the Pro plan?). Settings > Usage shows your next reset under Weekly limits.

The week resets at a fixed time: the pace line shows whether you are ahead of it

Example: 63% of the weekly limit used on day 4 of 7. The dashed line is a steady pace to the reset (Thu 18:00).

Example
Weekly usage against a steady pace63% of the weekly limit used after 4 days, where a steady pace would be at 57%. limit ≈ Thu 02:000%50%100%ThuFriSatSunMonTueWedThu ↺now63%limit ≈ Thu 02:00Weekly usage against a steady pace63% of the weekly limit used after 4 days, where a steady pace would be at 57%. limit ≈ Thu 02:000%50%100%ThuFriSatSunMonTueWed↺now63%limit ≈ Thu 02:00
  • Your usage
  • Steady pace
  • At the current rate

At this pacethe limit is reached around Thu 02:00, before the reset.

Weekly limit in this example
MeasureValue
Used so far63%
Days elapsed4 / 7
Steady pace today57%
At the reset, at this rate110%
Limit reachedThu 02:00
Weekly resetThu 18:00
Figure 2. Example values. The steady-pace line assumes even use across the 7 days; the reset day and time are fixed for your account.

Every message counts against the session and the week at the same time. The Claude Code docs warn that a single burst of heavy activity "can exhaust the weekly allowance before the session window resets" (Claude Code errors).

Fable and other model limits

Some models have a ceiling of their own inside the weekly limit (Claude Fable models on your plan):

  • Fable 5 and Fable 5.1 on Max and Ultra (and premium Team seats): up to 50% of your weekly limits at no extra cost, then usage credits or another model. It is a ceiling, not extra room: "you can never use more than your weekly limit".
  • Fable on Pro (and standard Team seats): not included in the plan's limits at all. It runs on usage credits from the first message.
  • Claude Code can also stop you with an Opus limit or a Sonnet limit that blocks only that model family; switching family with /model keeps you working (Claude Code errors).

A model limit can block you while the session and the week still have room

Example: session 35%, week 62%, Fable 100%. Whichever fills first stops you.

Example
Session, weekly and model limitsExample: session 35%, week 62%, Fable 100%. Whichever fills first stops you. Only Fable is blocked: switch model and keep working.35%sessionWeekly · all models62%Weekly · Fable only100%FullSession, weekly and model limitsExample: session 35%, week 62%, Fable 100%. Whichever fills first stops you. Only Fable is blocked: switch model and keep working.35%session5-hour session35%Weekly · all models62%Weekly · Fable only100%Full

Blockedonly Fable is blocked: switch model and keep working.

Limits in this example
LimitUsedResets
5-hour session35%—
Weekly · all models62%—
Weekly · Fable only100%—
Figure 3. Example values. Every message counts toward all of these limits at once; each one resets on its own schedule.

How much usage each plan gets

Anthropic doesn't publish limits in tokens or messages, only multiples of Pro: Max gives 5× Pro's usage every 5 hours and every week, Ultra 10× Pro's every week with no 5-hour limit once verified, and Pro "more usage than Free", with no multiplier (Claude pricing). Ultra was Max 20x until 30 October 2026: Anthropic says that 20x applied to the session limit, not to every limit, and the weekly limits didn't change (What is the Ultra plan?). On 22 September 2026, with Claude Opus 5.5, Anthropic raised the five-hour limits of Pro, Max (then both tiers), Team and seat-based Enterprise plans without saying by how much (Anthropic).

Anthropic publishes multiples of Pro, not token counts

Usage per week relative to Pro, as Anthropic describes its plans. Prices are not part of this chart.

Usage per week by planUsage per week relative to Pro: Pro the reference, Max 5× Pro, Ultra 10× Pro · no 5-hour limit.Prothe referenceMax5× ProUltra10× Pro · no 5-hour limitUsage per week by planUsage per week relative to Pro: Pro the reference, Max 5× Pro, Ultra 10× Pro · no 5-hour limit.Prothe referenceMax5× ProUltra10× Pro · no 5-hour limit

Per weekthe same multiplier buys fewer messages when each message carries a long conversation or big files.

Usage per week by plan
PlanRelative to Pro
Prothe reference
Max5× Pro
Ultra10× Pro · no 5-hour limit
Figure 4. Multipliers from Anthropic's pricing page (30 October 2026): Max gives 5× Pro's usage every 5 hours and every week, Ultra 10× Pro's every week with no 5-hour limit for verified accounts. Pro gives more usage than Free, without a figure; Free varies with demand.

Not sure which plan fits your use? The free Claude plan picker asks six questions and suggests a plan from Anthropic's published prices and multipliers.

PlanEvery 5 hoursEvery weekFablePrice (US)
Freelimited, no figurelimited, no figureno$0
Promore than Freemore than Freecredits only$20/month
Max (formerly Max 5x)5× Pro5× Pro≤ 50% of the week$100/month
Ultra (formerly Max 20x)no limit once verified10× Pro≤ 50% of the week$200/month

Source: claude.com/pricing, the Max plan article and the Ultra plan article, checked on 30 October 2026. Pro is $17 a month billed yearly; Max and Ultra are monthly only and include $100 and $200 a month of Claude API credits, which don't extend plan limits. Subscribers moved to the new names automatically, at the same price. Taxes excluded, local prices vary.

What fills a window

The unit is tokens. Anthropic lists what drives usage: message length, attachment size, the length of the current conversation, tools like Research and web search, the model, the effort level, artifacts and multi-step tasks (Usage limit best practices). The biggest one is invisible:

  • The whole conversation is sent again with every message. Anthropic's Claude Code docs say it outright: it "sends your full conversation with every request" (Manage costs). In a chat, a late reply costs far more than an early one, even for a one-line question.
  • Files and Project knowledge count while they are in context. Project content is cached and "counts less against your limits when reused", until the cache expires after a break (Usage limit best practices).
  • Model and effort. Bigger models and higher effort use more, and you can't turn off thinking on Opus 5.5 or the Fable models (Manage costs).
  • Tools and connectors are "token-intensive" (How usage and length limits work).
  • Claude Code also sends tool results (file contents, test output, diffs): "one debugging session can consume more than a day of chat" (Manage costs).

Each reply re-sends the whole conversation; a new chat starts cheap again

Tokens sent to the model at each turn of one chat (illustrative: 200-token messages, 600-token replies).

Illustrative
Tokens per turn in one conversationTurn 1 costs about 800 tokens. Every turn sends the whole history again, so turn 20 costs about 16k: 20 times more for a message of the same size.turn 136912151820≈ 800≈ 16k · ×20Tokens per turn in one conversationTurn 1 costs about 800 tokens. Every turn sends the whole history again, so turn 20 costs about 16k: 20 times more for a message of the same size.turn 18121620≈ 800≈ 16k · ×20
  • History sent again
  • Your message
  • Reply

Takeawaya new chat for each new topic brings the cost back to turn 1.

Tokens sent per turn in this illustration
TurnTokens sent
1800
21.6k
43.2k
64.8k
86.4k
108k
129.6k
1411.2k
1612.8k
1814.4k
2016k
Figure 5. An order of magnitude, not a measurement: caching makes re-read history count less, and real sizes depend on the model, the files and extended thinking.

Starting a new chat for each new topic is the cheapest habit there is.

What to do when a limit is full

Wait for the reset time Claude shows, use a limit reset if you have one (a Reset for free button in Settings > Usage, on the web and in Claude Desktop; subscribers received one on 22 September 2026), continue on usage credits on Pro, Max and Ultra, or switch model when only a model limit is full. A new chat or another browser doesn't help: the limits belong to your account. Each option is detailed in Claude usage limit reached: what it means and what to do.

How Claude's limits changed in 2026

Claude's limits changed eight times between March and October 2026

8 changes between March 2026 and October 2026. Dates in UTC.

Timeline of Claude's limit changesChanges to Claude's limits: Mar 13 → Mar 28: Promotion: double 5-hour usage outside weekday peak hours; Mar 26: Weekday peak hours (05:00–11:00 PT): 5-hour sessions drain faster; May 6: Claude Code 5-hour limits doubled; no peak drain for Claude Code on Pro and Max; May 13 → Sep 13: Promotion: Claude Code weekly limits +50%; Jul 20: Fable 5 in Max at up to 50% of weekly limits; on Pro, usage credits only; Sep 14: Claude Code weekly limits permanently +25% over the pre-promotion level; Sep 22: Claude Opus 5.5: 5-hour limits raised on Pro, Max and Team; one limit reset for subscribers; Oct 30: Max 5x renamed Max, Max 20x renamed Ultra: no 5-hour limit on verified Ultra accounts, 10x Pro's weekly usage.MarAprMayJunJulAugSepOct12345678Timeline of Claude's limit changesChanges to Claude's limits: Mar 13 → Mar 28: Promotion: double 5-hour usage outside weekday peak hours; Mar 26: Weekday peak hours (05:00–11:00 PT): 5-hour sessions drain faster; May 6: Claude Code 5-hour limits doubled; no peak drain for Claude Code on Pro and Max; May 13 → Sep 13: Promotion: Claude Code weekly limits +50%; Jul 20: Fable 5 in Max at up to 50% of weekly limits; on Pro, usage credits only; Sep 14: Claude Code weekly limits permanently +25% over the pre-promotion level; Sep 22: Claude Opus 5.5: 5-hour limits raised on Pro, Max and Team; one limit reset for subscribers; Oct 30: Max 5x renamed Max, Max 20x renamed Ultra: no 5-hour limit on verified Ultra accounts, 10x Pro's weekly usage.MAMJJASO12345678
  1. Mar 13 → Mar 28Promotion: double 5-hour usage outside weekday peak hours
  2. Mar 26Weekday peak hours (05:00–11:00 PT): 5-hour sessions drain faster
  3. May 6Claude Code 5-hour limits doubled; no peak drain for Claude Code on Pro and Max
  4. May 13 → Sep 13Promotion: Claude Code weekly limits +50%
  5. Jul 20Fable 5 in Max at up to 50% of weekly limits; on Pro, usage credits only
  6. Sep 14Claude Code weekly limits permanently +25% over the pre-promotion level
  7. Sep 22Claude Opus 5.5: 5-hour limits raised on Pro, Max and Team; one limit reset for subscribers
  8. Oct 30Max 5x renamed Max, Max 20x renamed Ultra: no 5-hour limit on verified Ultra accounts, 10x Pro's weekly usage
  • Promotion
  • Less usage
  • More usage
  • Model rules
Figure 6. Dates in UTC, from Anthropic's announcements and help center; the text cites each source.

The dates come from Anthropic: the March promotion page (archived), Thariq's peak-hours thread, the 6 May announcement, @claudeai on Fable, the Claude Code weekly limits promotion page, the Opus 5.5 announcement and the Ultra plan article. Before 30 October 2026, "Max" covers both tiers of the time, Max 5x and Max 20x, now Max and Ultra. Peak hours by time zone, and what they still change: Claude peak hours.

Every limit at a glance

LimitPlansWhen it resetsWhere to see it
5-hour sessionFree, Pro and Max (Ultra: only until verified)about five hours after it openedSettings > Usage (Current session), /usage in Claude Code
Weekly, all modelsevery plan (Free: no figure)on paid plans, a fixed day and time assigned to your accountSettings > Usage (Weekly limits), /usage
Fable ceilingMax, Ultra, premium Team seatswith the weekly limitSettings > Usage, under Weekly limits
Opus or Sonnet limitClaude Code, when your plan has onethe time in the messagethe Claude Code message, /usage
Usage creditsPro, Max and Ultra, if you turn them onmonthly, against the spend limit you setSettings > Usage (Usage credits)

Where each view lives, and what it misses, is in how to check your Claude usage.

How UsageMeter helps

Settings > Usage is accurate, but you have to open it. UsageMeter, a free extension for Chrome, Edge and Firefox (independent, not affiliated with Anthropic), shows every limit claude.ai reports for your account, each with its reset countdown and a pace marker that says whether you are on track: the session in the composer (hover it for the rest) and the limit you choose on the toolbar icon. Its popup adds this session's tokens and how your week split between chats, Claude Code and Cowork. Alerts before a limit is full and when it comes back, and charts, come with Pro (compare plans).

Weekly limits card: all models at 63%, on track, with today's budget; Fable at 87%, active, with the limit expected before the reset (sample data)Données d'exemple
Chaque limite hebdomadaire avec sa réinitialisation, et le repère de rythme : au-delà, vous consommez la semaine plus vite qu'un rythme régulier.

FAQ

Does Claude's 5-hour limit reset at a fixed time?

No. It is a rolling window: it closes about five hours after the message that opened it, and the next one opens with your next message. Settings > Usage shows the exact time remaining. Only a limit reset refills it early: Anthropic grants them occasionally, and gave subscribers one on 22 September 2026. A verified Ultra account has no 5-hour limit at all.

Is Claude's weekly limit a rolling seven days?

No. Anthropic says the weekly limit resets at a fixed time each week assigned to your account, and that the day and time don't depend on when you started using Claude or subscribed. Settings > Usage shows your next reset under Weekly limits.

Do Claude Code and claude.ai share the same limits?

Yes. Claude on the web, Claude Desktop, the mobile apps, Cowork and Claude Code all count toward the same session and weekly limits. Claude Code can also hit an Opus or Sonnet limit that blocks only that model family, while other models keep working.

How many messages does a Claude plan give you?

Anthropic doesn't publish a number, because it depends on conversation length, attachments, model, effort and tools. It publishes multiples of Pro instead: Max gives 5× Pro's usage every 5 hours and every week, Ultra 10× Pro's every week with no 5-hour limit once verified, and Pro "more usage than Free" (Claude pricing FAQ).

Do Claude's limits change at peak times?

Anthropic said on 26 March 2026 that weekday peak hours (05:00 to 11:00 Pacific Time) make the 5-hour session drain faster on Free, Pro and Max, and it hasn't announced an end for chat. It removed that reduction for Claude Code on Pro and Max on 6 May 2026. Weekly limits were never affected.

Sources

What changed

  • 30 October 2026: Max 5x and Max 20x renamed Max and Ultra; Ultra has no 5-hour limit once Anthropic has verified the account and 10× Pro's weekly usage; the weekly multiples (Max 5×, Ultra 10×) and Free's weekly limit (no figure) added, with the API credits of Max and Ultra; Pro's "at least 5× Free" removed (no longer published); the change added to the 2026 timeline. Weekly limits unchanged.
  • 24 September 2026: new guide, every fact checked against the sources above: weekly reset at a fixed weekly time, Fable ceiling, limit resets, usage credits (formerly Extra usage), the 22 September 2026 increase of five-hour limits, 2026 timeline.

UsageMeter

UsageMeter est une extension indépendante qui affiche vos limites d'utilisation Claude sur claude.ai, avec le compte à rebours de chaque réinitialisation. Non affiliée à Anthropic.

Rédigé avec l'aide de Claude, vérifié par UsageMeter le 30 octobre 2026. Comment nous écrivons