If you use Claude Code through a Claude subscription, Anthropic won't tell you how many messages or tokens you get. What it does publish is the shape of the allowance: a session limit that resets every five hours, a weekly limit that resets once a week, and a set of multiples that show how the plans compare with one another. Both limits are shared with Claude on the web, desktop and mobile, so a long chat in the browser draws on the same budget as a coding session in your terminal. Anthropic says what any single request costs depends on the length and complexity of the conversation, the features you use, the model and the effort level, which is why it describes allowances instead of counts.

Everything here comes from Anthropic's help center, its pricing page, the Claude Code documentation, the Anthropic news blog and posts from the official Claude and ClaudeDevs accounts on X, checked on October 17, 2026, Beijing time. Anthropic adjusts these rules often, so we'll keep updating this page as it does.

Two limits and one shared pool

According to Anthropic's help center, every paid plan has two usage limits running at once. The session limit resets every five hours. The weekly limit applies across all models and resets at a fixed time each week that Anthropic assigns to your account; that day and time stay the same no matter when you start using Claude or when your subscription began, and you can see the next reset in Settings > Usage. The Claude Code documentation adds that usage counts against both allowances at the same time, so a single burst of heavy work, such as a large workflow fanout, can use up the weekly allowance before the five-hour window has even reset.

The pool covers more than the terminal. Anthropic says usage on claude.ai, Claude Code and Claude Desktop all counts toward the same limit, and Claude Code running inside VS Code, Cursor or JetBrains IDEs counts too. The Pro and Max help articles also note that Anthropic may limit usage in other ways, such as weekly and monthly caps or limits on particular models and features, at its discretion.

To see where you stand, Anthropic points to the progress bars in Settings > Usage. Inside Claude Code, the /usage command shows your plan usage bars along with a breakdown that attributes recent usage to skills, subagents, plugins and individual MCP servers, and flags behaviors such as long context or cache misses when they account for a meaningful share. Claude Code also warns you before a window runs out, with messages such as "You've used 85% of your session limit · resets 3:45pm."

What each plan gets

Anthropic describes plan sizes as multiples, not quantities. This is how its pricing page and help center lay out Claude Code access today:

  • Free: Claude Code isn't included.
  • Pro ($20 a month, or $17 a month billed annually at $200): includes Claude Code, a five-hour session limit and a weekly limit. Anthropic's pricing page says Pro gives you "at least 5x more usage per 5-hour session than Free." Fable 5 and Fable 5.1 aren't part of Pro's included usage and run only on usage credits.
  • Max 5x ($100 a month) and Max 20x ($200 a month): five and 20 times the Pro plan's per-session usage allowance, plus a weekly limit across all models and a separate weekly limit for Fable. Max subscribers can spend up to 50% of their weekly limits on Fable models.
  • Team: Claude Code comes with every seat. Standard seats cost $20 per seat per month billed annually or $25 monthly and get "more usage than Pro." Premium seats cost $100 billed annually or $125 monthly and get five times the usage of a Standard seat, and they can put up to 50% of their weekly limit toward Fable models, which standard seats can only use with usage credits. Each seat's allowance resets on a rolling five-hour window and a weekly window, shared with Claude chat and Cowork.
  • Enterprise: on usage-based Enterprise plans there are no per-seat usage limits, and usage is billed at API rates. The pricing page lists these at $20 per seat per month billed annually, with a 20-seat minimum, plus usage. Legacy seat-based Enterprise plans have seat allowances similar to Team.

Two details are easy to misread. Anthropic's Max article frames the 5x and 20x multiples against Pro's per-session allowance and doesn't say how the weekly limits compare. And the Fable allowance on Max isn't extra capacity: Anthropic's own FAQ answers "No" to whether Max users get 50% more for Fable, explaining that Fable can take up to half of the weekly limit while every other model draws from the same pool. Max plans also come with $100 or $200 a month in Claude API credits, but Anthropic says those credits can't be used in Claude Code, including for extra usage.

Are weekly limits 50% higher?

Not anymore, though they were. Anthropic's help center article on the promotion says that from May 13 through September 13, 2026, weekly usage limits in Claude Code were 50% higher for Pro, Max and Team plans and for legacy seat-based Enterprise users. Free plans and consumption-based Enterprise seats were excluded, five-hour limits were unaffected, and the boost applied only to Claude Code, in the CLI, IDE extensions, desktop and web, not to the Claude apps or Cowork. ClaudeDevs first announced it on May 14, Beijing time, and Anthropic extended it four times before it ended at 11:59 p.m. Pacific on September 13, which was 14:59 on September 14 in Beijing.

What replaced it is a smaller permanent increase. The same article now says that "starting September 14, 2026, weekly limits in Claude Code are 25% higher than they were before the promotion" for Pro, Max, Team and seat-based Enterprise plans. When ClaudeDevs announced the change on August 30, Beijing time, it added that compared with the promotional level, "this works out to a 17% reduction in weekly limits on Claude Code." Any page that still says weekly limits are 50% higher is describing the promotion that ended.

The five-hour limit has moved since then. In its Claude Opus 5.5 announcement, Anthropic said it is increasing five-hour usage limits on Pro, Max, Team and seat-based Enterprise plans. ClaudeDevs put the Claude Code increase at 20% in a post at 00:44 on September 23, Beijing time, and said that because Opus 5.5 is priced lower than Opus 5, "your 5-hour and weekly limits go 25% further." Earlier, on August 8, ClaudeDevs said the classifier that auto mode runs on every tool call no longer counts toward usage limits on Pro, Max and Team plans.

How resets work

The regular cycle is simple: the session limit resets every five hours, and the weekly limit resets at your account's fixed time. If you upgrade, Anthropic says the new plan's limits apply right away.

On top of that, Anthropic occasionally hands out limit resets to eligible plans. Depending on the reset you're offered, it puts either your five-hour session limit or your weekly limit back to full the moment you use it, and you don't have to wait until you hit a limit. Your weekly limit still resets on its usual day and time afterward. A reset can't be undone, any expiry date appears in Settings > Usage, and it doesn't refund extra usage you've already been billed for. If you downgrade or cancel before using it, it's gone. The "Reset for free" button lives in Settings > Usage on the web and in Claude Desktop, and on the message you see when you reach a limit, but not in Claude Code in the terminal or IDE. Because limits are shared, a reset applied on the web restores Claude Code too. Anthropic's help center doesn't describe any way to buy a reset.

The most recent one came with the Opus 5.5 launch. ClaudeDevs said Pro, Max and Team users could find the reset in Settings > Usage and apply it any time until October 22. Anthropic has also reset everyone's limits outright from time to time; the latest such post from ClaudeDevs came with the Fable 5.1 release at 02:35 on September 2, Beijing time.

Claude Code also tries not to stop mid-edit. If you reach your five-hour limit while it is partway through a response, Anthropic says it may keep working briefly to finish the current step, showing "Usage limit reached · wrapping up." That wrap-up allowance is small and capped, it counts toward your weekly limit, and it only covers a response already in progress. On Pro it applies up to once per weekly usage period; on Max it can apply each time you hit a five-hour limit, within your weekly limit. Anthropic says it is rolling out gradually. Since version 2.1.234, interactive sessions signed in with a subscription also wait for the reset by default and pick the task back up, showing "Usage limit reached · limit resets 3:45pm" and "Continuing automatically at 3:45pm · esc to cancel." You can turn this off in /config.

What the limit messages mean

People often search for a "rate limit reached" error, but that isn't the wording Anthropic's documentation uses for plan limits. Claude Code shows different messages for different ceilings, and they don't all mean the same thing:

  • "You've hit your session limit · resets 3:45pm" or "You've hit your weekly limit · resets Mon 12:00am": your plan allowance has run out. Both limits are shared across all models, so switching models won't restore access. In the Claude apps the equivalents are "5-hour limit reached - resets [time]" and "You've reached your weekly limit. It resets [day] at [time]."
  • "You've hit your Opus limit" or "You've hit your Sonnet limit": these apply only to that model family, and the docs say switching to a model outside it with /model keeps you working.
  • "Server is temporarily limiting requests (not your usage limit)": a short-lived server-side throttle that has nothing to do with your plan quota. Claude Code retries it automatically.
  • "Request rejected (429)": you've hit the rate limit configured for an API key, Amazon Bedrock project or Google Cloud project. The docs suggest running /status first, because a stray ANTHROPIC_API_KEY in your environment can route requests through a low-tier key instead of your subscription.
  • "Repeated 529 Overloaded errors": the API is at capacity. Anthropic says a 529 is not your usage limit and doesn't count against your quota.

Paying past the limit with usage credits

On Pro and Max, you can switch on usage credits in Settings > Usage, or with the /usage-credits command in Claude Code. Once one of your limits runs out, requests move to pay-as-you-go billing at standard API rates and keep going until you reach the monthly spend limit you set or your prepaid balance runs out with auto-reload off. The warning about approaching your limit can linger for a moment, but Anthropic says it changes to "Now using credits" once the next request goes through. Your five-hour and weekly limits keep resetting on their usual schedule.

Anthropic also sells prepaid usage bundles at a discount: $45 buys $50 of usage, $200 buys $250 and $700 buys $1,000, and Pro and Max subscribers can buy up to $2,000 worth of discounted bundles a month. In most places usage credits don't expire, though Anthropic says that in some jurisdictions, such as Japan, credits bought from September 10, 2026 expire after six months. On Team and seat-based Enterprise plans, admins decide whether members can use credits. One more case is an entitlement check rather than a quota: when a model's 1M-token context is only included through usage credits on your plan, Claude Code shows "Usage credits required for 1M context," even if your session and weekly allowances still have room.

The API key route

You can also skip plan limits entirely by signing in with an API key from the Claude Console, Amazon Bedrock, Google Vertex AI or Microsoft Foundry. Anthropic says that usage is billed per token to that account, with no hard stop; the account is simply charged for what it uses. Claude subscriptions and the Console are separate products, and the Pro plan doesn't include any API usage. Anthropic's Claude Code cost guide says that across enterprise deployments, the average is around $13 per developer per active day and $150 to $250 per developer per month, with 90% of users staying below $30 per active day.

There's one trap Anthropic warns about. If an ANTHROPIC_API_KEY environment variable is set, Claude Code uses that key instead of your subscription, and you're billed at API rates even while logged in with your Claude account. Anthropic recommends keeping the variable unset unless you need it and running /status to check which authentication method is active.

How Claude Code compares with Codex

Codex works on a similar two-window model, and our explainer "OpenAI Keeps Codex's Weekly Caps Vague: What Each ChatGPT Plan Actually Gets" covers it in full as of our October 9 check. The differences are mostly in how much each company discloses and what you can buy. OpenAI publishes ranges of local messages per five-hour window for Plus and Business Standard, while Anthropic publishes only plan multiples. OpenAI says a new five-hour Codex window starts with your first message after the previous one ends; Anthropic says your Claude weekly limit resets at a fixed time assigned to your account. Both share their allowance with other products, ChatGPT Work on one side and the Claude apps on the other. When you run out, Codex offers credits, paid instant resets on some accounts and free banked resets, while Claude offers usage credits at API rates and occasional free limit resets. Both let you move to API billing with an API key. OpenAI's resets, many of them announced by Tibo, are logged on our "Tibo Codex Reset Tracker: Latest Announcements and History" page, which doesn't cover Anthropic.

Making the allowance last

Anthropic's own advice centers on model choice and context. Its help center says Opus uses meaningfully more of your quota than Sonnet, recommends Sonnet for most coding work and Haiku for quick mechanical tasks, and suggests planning with Opus and executing with Sonnet, which Claude Code builds in as /model opusplan. Because every turn resends the conversation so far, Anthropic calls /clear between unrelated tasks the single most effective lever for quality and cost, with /compact for long tasks you're still in the middle of. It also suggests pointing Claude at file paths instead of pasting files, keeping CLAUDE.md to roughly 200 lines, and asking for a plan before changes that touch more than a few files. If none of that is enough and you keep hitting the weekly wall, Anthropic's path is a higher tier: Max 5x from Pro, or Max 20x from Max 5x.