Short answer
Start on Pro ($20). Move to Max 5x ($100) when Pro's five-hour limit stops you in most sessions, and to Max 20x ($200) only if Max 5x still does. Max gives you more usage per session, on the same Opus and Sonnet. At Anthropic's $13 enterprise average day, API billing beats Pro only below 1.5 active days a month.
Claude Code pricing by plan, as of September 2026
Claude Code is included in every paid Claude plan: Pro, Max, Team and Enterprise. Here's what Anthropic's pricing page, the Max article and the Pro article say.
- Pro: $20 a month, or $17 a month on the annual plan ($200 billed up front). The baseline that the Max multiples are measured against.
- Max 5x: $100 a month, monthly only. 5x Pro's usage per five-hour session.
- Max 20x: $200 a month, monthly only. 20x Pro's usage per five-hour session.
- Weekly limit: all three have one. Anthropic doesn't publish the size of Pro's, or a weekly multiple for Max.
- Fable 5.1: usage credits only on Pro. Included on Max up to 50% of weekly limits.
- Claude Code: included in all three.
Prices exclude tax, and Anthropic says prices and plans can change at its discretion. The Max prices are for web subscriptions, and mobile pricing may vary by app platform. Max also lists higher output limits, early access to advanced features and priority access at high traffic times. The pricing table shows priority access under Max only, while the Pro article lists it for Pro too. So all that row tells you is that Max has it.
A Team seat is $25 a month ($20 a month on the annual plan) for Standard, with 1.25x Pro's per-session allowance. It's $125 ($100 a month on the annual plan) for Premium, with 6.25x. Teams run from 2 to 150 members, and the limits are per member (Team plan article). A Premium seat isn't a Max plan. You buy it as a seat on a Team plan with at least 2 members, and its 6.25x sits between Max 5x and Max 20x. Enterprise is $20 per seat a month billed annually, plus usage at API rates. Claude Code is included in the single seat on new and self-serve Enterprise plans (Claude Code for Team and Enterprise).
Which plan fits which workload
Anthropic doesn't publish how many hours Pro covers, so you have to go by what you see yourself. The rule below is ours. Anthropic didn't write it.
- Stay on Pro if the five-hour limit rarely stops you and the weekly limit has never ended your week early.
- Move to Max 5x if Pro's five-hour limit stops you in most working sessions. Say Pro cuts you off two hours into a window: five times that is about ten hours, more than the whole five-hour window.
- Move to Max 20x only if Max 5x still stops you inside a window. Say Pro cuts you off 30 minutes in. Max 5x covers about two and a half hours, and Max 20x a full five-hour window.
The weekly limit is its own wall. Anthropic gives the Max multiples per five-hour session only, and no Anthropic page states a weekly multiple. If Pro's weekly limit is what stops you, the published figures can't tell you how much more week Max gives. Check /usage after your first full week.
Treat the hours above as arithmetic on Anthropic's published multiples, with no promise that usage scales in a straight line. Everything you do while signed in to your account draws on the same pool. That's chat on the web, desktop and mobile, plus every Claude Code session you run at once (pricing FAQ).
Measure it yourself. /usage in Claude Code shows your plan usage bars and when they reset. The Usage page in your Claude account settings shows the session limit and the weekly limits (Claude Code docs, help article).
Watch the dates, though. Claude Code's weekly limits changed on 14 September 2026, when a 50% promotion ended and they settled at 25% above where they were before it. Anthropic raised the five-hour limits on 22 September 2026. Any reading of where Pro stops you that you took before those dates is out of date. The Claude Code plan calculator applies this rule to your own hours and Pro reading, rates each plan, and tells you when it can't.
What Anthropic does not publish
A lot of Claude plan advice online quotes numbers Anthropic doesn't currently publish. When we last checked, no Anthropic page stated any of these:
- How many messages or tokens a five-hour window or a week holds on any plan. The pricing FAQ says there's no fixed message count.
- Pro's own allowance. Only multiples of Pro are published: 5x and 20x for Max, and 1.25x and 6.25x for Team seats.
- Any weekly multiple for Max 5x or Max 20x. The 5x and 20x are stated per five-hour session.
- What any limit is in hours or tokens, including how big the 22 September 2026 five-hour increase was.
- What opens the five-hour window. Anthropic calls it a rolling window. That your first message opens it is widely reported, and Anthropic doesn't document it.
- How much faster Opus uses your quota than Sonnet. Anthropic says Opus uses meaningfully more quota and costs several times more per turn, and gives no figure for quota.
- How "up to 50% of weekly limits" for Fable converts to tokens, or how long that rule will last.
- Whether the paused credit for Agent SDK and
claude -pusage will come back, or when.
When a page gives you any of these as a fact, it's an estimate or a guess. Ours are labelled as estimates, with the assumption next to them.
API key or subscription: the break-even
Claude Code can bill two ways. A plan charges a flat price and enforces the limits above. An API key charges per token at list price, with no plan limits. Anthropic's planning figures for API-billed Claude Code are about $13 per developer per active day and $150 to $250 per developer per month across enterprise deployments (Claude Code docs). Costs stay below $30 per active day for 90% of users. Those are API bills, so they tell you what the same work costs without a plan.
The break-even is one division:
Break-even days at Anthropic's $13 enterprise average day, and at the $30 day that 90% of users stay below:
| Plan | At $13 a day | At $30 a day |
|---|---|---|
| Pro ($20) | 1.5 days | 0.7 days |
| Max 5x ($100) | 7.7 days | 3.3 days |
| Max 20x ($200) | 15.4 days | 6.7 days |
If you use Claude Code on more active days a month than the number in the table, the plan costs less than the same tokens at API list price, as long as the plan's limits cover that usage. At Anthropic's enterprise average day, API billing beats Pro only below about 1.5 active days a month. It beats Max 5x below about 7.7.
A worked example: 10 active days a month at $13 a day is $130 on the API. That's more than Max 5x's $100 and far more than Pro's $20, so a plan that covers your sessions costs less than the API. It's less than Max 20x's $200, so Max 20x would cost more than the API for the same 10 days.
Your own day may cost more or less than the average. The cost figure in /usage estimates what a session would cost at API list price. Anthropic says it isn't relevant to billing on Pro and Max, but it's a fair number to use as your API cost per active day. For your own hours and model mix, the Claude Code plan calculator prints an estimate, labelled as one, with its assumptions.
API list prices per million tokens, as of September 2026 (Claude API pricing):
| Model | Input | Output | Cache read | Cache write (5 minutes) |
|---|---|---|---|---|
| Fable 5.1 | $10 | $50 | $0.25 | $12.50 |
| Opus 5.5 | $4 | $20 | $0.20 | $5 |
| Sonnet 5.5 | $2 | $10 | $0.20 | $2.50 |
| Haiku 4.5 | $1 | $5 | $0.10 | $1.25 |
Anthropic says Haiku 5.5 will join the Claude 5.5 family in the coming weeks, so recheck the Haiku row before you rely on it.
Most of a Claude Code bill is cache reads: Anthropic says they make up the majority of the cost of agentic coding work. In its data from March to September 2026, Claude reads 324 input tokens for every output token, up from 189 (Anthropic, 24 September 2026). On a plan the prompt cache lasts an hour. On an API key it lasts five minutes by default (Claude Code docs).
One trap. If ANTHROPIC_API_KEY is set in your environment, Claude Code uses it instead of your plan and bills API usage. Interactive sessions ask once whether to approve the key, and claude -p uses it whenever it is set. To go back to your plan:
To stop being offered API credits at a limit, sign in with only your plan credentials: claude auth logout, then claude auth login.
For API bills from agents that aren't coding CLIs, see The real cost of running a persistent AI agent in 2026.
What happens when you hit a limit, and what to do
Claude Code names the limit and its reset time. The four messages, from Anthropic's error reference:
The session and weekly limits cover every model, so switching models doesn't help. The Opus and Sonnet limits cover only that model family. Run /model to pick a model outside it and you keep working, though the first request after the switch misses the prompt cache. A single burst of heavy activity can use up the weekly allowance before the session window resets.
- Find out which limit it is.
/usageshows your plan usage bars and when they reset./rate-limit-optionsopens the menu of choices. To estimate when a five-hour window closes, the Claude Code limit reset calculator works from the time of your first prompt, which is how the window is widely reported to open. - Wait in the same session. Claude Code 2.1.234 (17 August 2026) and later waits in an open interactive session signed in with your plan, then continues the task after a usage limit resets (docs). If the computer slept through the reset for more than about 30 minutes, press Enter to continue. It doesn't start on its own in Remote Control sessions, or when the reset is more than 24 hours away (often the case for a weekly limit). You can still start a wait from
/rate-limit-options, and it keeps counting down.claude -pand background sessions don't get it. - Use usage credits.
/usage-creditsmanages them. They're prepaid and billed at standard API rates, separately from your plan, with an optional monthly spend cap and auto-reload (Anthropic's article). Bundles are sold at 10%, 20% or 30% off: $50 of credits for $45, $250 for $200, and $1,000 for $700 (article dated 18 May 2026). - Switch to a Claude Console account for API-billed work during an intensive sprint (Anthropic's article).
- Move up a tier. Pro to Max 5x, or Max 5x to Max 20x. On Max 20x, Anthropic's options are usage credits, a Console account or waiting.
- Use a limit reset if you have one. Anthropic occasionally offers a reset that puts the five-hour or weekly limit back to full. It gave subscribers a saveable one when Opus 5.5 launched on 22 September 2026. Open Settings, then Usage, and use "Reset for free" on Claude on the web or desktop. The button isn't in Claude Code or on mobile, but the reset applies everywhere because limits are shared, and weekly limits still reset on their usual day (Anthropic's article).
The wait ends if you exit Claude Code or hand the session to another surface, and it re-arms at most twice in a row.
Agent SDK and claude -p usage: Anthropic announced a separate monthly credit for it, then paused that on 15 June 2026. Its help article, dated 16 June 2026, says usage from the Agent SDK, claude -p and third-party apps still draws from your plan's limits. It also says Anthropic will announce any change before it takes effect (Anthropic's article). We haven't found anything newer, so recheck before you build on it.
Model choice changes how fast the quota goes
The default changed. Opus 5.5 has been Claude Code's default on Pro, Max, Team and Enterprise since 22 September 2026 (Claude Code 2.1.280). Before that, Pro and Team Standard defaulted to Sonnet (Claude Code docs).
Anthropic's guide says Sonnet is the right choice for the large majority of coding work, and that Opus uses meaningfully more of your quota (Anthropic's guide). The default is Opus, so switching models is how you save quota: /model picks a model, and /model opusplan plans with Opus and executes with Sonnet.
Fable is never the default. Max plans include Fable 5 and 5.1 up to 50% of their weekly limits, and the Fable models draw the same pool faster. Pro uses usage credits for them (Anthropic's article).
Thinking can't be turned off on Opus 5.5, Sonnet 5.5 or the Fable models, and thinking tokens bill as output. Lowering /effort keeps that cost down. Fast mode is a research preview that runs Opus 5.5 up to 2.5x faster at $8 input and $40 output per million tokens. On a plan it runs on usage credits only and doesn't count against your plan's limits (docs).
When Anthropic's own cloud is enough
Anthropic will run Claude Code in the cloud for you. A cloud session runs on Anthropic-managed machines, so it keeps going while your laptop is shut. It's available on Pro, Max and Team plans, and to Enterprise users on premium or Chat + Claude Code seats. It shares your plan's limits, and Anthropic says there's no separate compute charge for the cloud machine (Claude Code docs).
It fits tasks on GitHub repositories that can run unattended. Cloning a repository and opening pull requests need GitHub. A repository hosted elsewhere can be sent as a local bundle, but the results can't be pushed back. By default you work in Anthropic's environment. You don't get a computer of your own.
With Remote Control you drive a session from your phone or browser, but it runs on your own machine, which has to stay on with the claude process running. It reconnects after the machine sleeps, and it doesn't work with API keys (Claude Code docs).
When you sign in with your plan, every one of these options draws on the same plan limits. A cloud session, Remote Control, a VPS and a Hivra computer all spend your account's five-hour and weekly allowance. None of them raises it. On an API key there's no plan allowance to raise. To compare the ways of keeping a run going with the lid shut, see Will Claude Code keep running if I close my laptop?
Running Claude Code off your laptop on Hivra
If you want a computer of your own that stays on, Hivra runs Claude Code on one, and you sign in with your own Anthropic account. The work runs there instead of on your laptop. Your five-hour and weekly limits follow your account, so a Hivra computer doesn't raise them.
Hivra plans are $9.99 a month for 2 vCPU and 4 GB of RAM, or $19.99 a month for 4 vCPU and 8 GB of RAM, with a 7-day money-back guarantee on card payments. Hivra runs the official Claude Code CLI, and on your own Anthropic login it adds no markup on Claude usage. That usage bills through Anthropic against your plan.
On a paid Hivra plan the computer stays on and keeps your files, sessions and login. A Claude Code or Codex run you start inside tmux in the computer's Terminal tab keeps going after you close the laptop. On Claude Code, work you send through Telegram, connected in the agent's Telegram tab, runs on the computer, not in your browser.
Anthropic sets the terms for hosting Claude Code. Its legal page says the advertised usage limits for Pro and Max assume ordinary, individual usage of Claude Code and the Agent SDK. Read it before you put a plan behind round-the-clock automation.
Hivra isn't always the better choice. If your tasks live in GitHub repositories and can run unattended, Anthropic's cloud sessions have no separate compute charge and draw on your plan's limits. If a run only has to keep going while the laptop stays open and plugged in, a keep-awake command is enough, and the keep-awake command builder writes it. A computer of your own makes sense when you want your files, sessions and login to stay in place between tasks.
Start Claude Code from the Claude Code agent page. Hivra is independent and is not affiliated with Anthropic or OpenAI.
What changed and when: the Claude plan changelog
Dated changes to Claude plans and Claude Code limits, each with the Anthropic page that states it. Anthropic changes plans often, so open the linked page before you rely on a row. Changes that only news sites or posts on X reported are left out.
- 28 September 2026: Sonnet 5.5 launched at $2 and $10 per million tokens, the same price as Sonnet 5. Source: Claude Sonnet 5.5 announcement.
- 22 September 2026: Claude Code made Opus 5.5 the default model, including on Pro and Team Standard plans, where Sonnet had been the default (v2.1.280). Source: Claude Code changelog.
- 22 September 2026: Opus 5.5 launched at $4 and $20 per million tokens. Anthropic raised five-hour usage limits on Pro, Max, Team and seat-based Enterprise plans and gave subscribers a rate limit reset they can save and use when they choose. It did not publish the size of the increase. Source: Claude Opus 5.5 announcement.
- 14 September 2026: Weekly limits in Claude Code became 25% higher than they were before the May to September promotion, on Pro, Max, Team and seat-based Enterprise plans. Claude Code only. Source: Claude Code May to August 2026 weekly limits promotion.
- 1 September 2026: Fable 5.1 launched. Source: Claude release notes.
- 17 August 2026: Claude Code v2.1.234 added waiting out a usage limit in the open session and continuing after the reset. Source: Claude Code changelog.
- 24 July 2026: Opus 5 launched. Source: Claude release notes.
- 19 July 2026: The Fable 5 promotion ended at 11:59:59 PM PT. Since then Max plans and premium seats include Fable up to 50% of their weekly limits, and Pro and Team Standard plans use it on usage credits. Source: Claude Fable models on your plan.
- 30 June 2026: Sonnet 5 launched and became Claude Code's default model (v2.1.197). Source: Claude Code changelog.
- 15 June 2026: Anthropic paused a planned separate monthly credit for Agent SDK and claude -p usage. That usage still draws from your plan's limits, and Anthropic says it will announce any change before it takes effect. Source: Use the Claude Agent SDK with your Claude plan.
- 19 May 2026: Claude Code renamed extra usage to usage credits, and the /extra-usage command became /usage-credits (v2.1.144). Source: Claude Code changelog.
- 13 May 2026: A temporary 50% increase to weekly limits in Claude Code began on Pro, Max, Team and seat-based Enterprise plans. It ran through 13 September 2026. Source: Claude Code May to August 2026 weekly limits promotion.
- 6 May 2026: Claude Code's five-hour limits doubled on Pro, Max, Team and seat-based Enterprise plans, and the peak-hours limit reduction on Claude Code was removed for Pro and Max. Source: Higher Claude Code limits announcement.
- 12 February 2026: Self-serve Enterprise plans launched with one seat type covering Claude and Claude Code. Source: Claude release notes.
- 28 January 2026: Team prices were lowered to $20 a month on the annual plan or $25 a month on the monthly plan for a Standard seat, and $100 a month on the annual plan or $125 a month on the monthly plan for a Premium seat, with Claude Code in every seat. Source: Claude Team updates.
- 16 January 2026: Claude Code was added to Team Standard seats. Source: Claude release notes.
- 3 December 2025: Pro users got Opus 4.5 in Claude Code as part of their subscription (v2.0.58). Source: Claude Code changelog.
- 9 April 2025: Max launched at $100 a month for 5x Pro's usage and $200 a month for 20x. Source: Max plan announcement, archived copy.
Common questions
Is Claude Code included in Pro?
Yes. Claude Code is included in Pro, Max, Team and Enterprise, and can also be billed to an API key instead of a plan. Pro is $20 a month billed monthly, or $17 a month on the annual plan ($200 billed up front). Pro doesn't include API usage through the Claude Console. Prices from Anthropic's pricing page, as of September 2026.
What is the Claude Max plan?
Max is Anthropic's higher-usage individual plan, in two sizes: Max 5x at $100 a month and Max 20x at $200 a month, billed monthly only. You get five or twenty times Pro's usage per five-hour session, plus Claude Code and the Fable models up to 50% of Max's weekly limits. Anthropic also lists higher output limits, early access to advanced features and priority access at high traffic times. Prices as of September 2026, before tax.
What are Claude Max limits?
Anthropic states them as multiples of Pro's usage per five-hour session: Max 5x is five times and Max 20x twenty times. Both also have weekly limits, which apply across all models and reset at a fixed time assigned to your account. Anthropic doesn't publish a weekly multiple, message or token counts, or a size for Pro's own allowance. So any weekly multiple or hour count you see elsewhere isn't one Anthropic currently publishes.
What is the difference between Claude Pro and Max?
Price, capacity and a few extras. Pro is $20 a month, or $17 a month on the annual plan. Max 5x is $100 and Max 20x is $200, both billed monthly only. The capacity gap is stated per five-hour session, at five and twenty times Pro's. Max also lists higher output limits, early access to advanced features and priority access at high traffic times. It includes Fable models up to 50% of its weekly limits, where Pro pays usage credits for them.
Is Max worth it for Claude Code?
Only if Pro stops you. If Pro's five-hour limit stops you in most sessions, Max 5x at $100 a month gives five times the per-session allowance, and Max 20x at $200 gives twenty times. If Pro rarely stops you, Max is money you don't need to spend. Check /usage after your first full week to see how the weekly limit treats you.
Is Claude Pro enough for Claude Code?
Yes, if Pro's limits cover your sessions. Anthropic doesn't publish how many hours that is. Claude Code is included in Pro at $20 a month, and Pro runs both Opus and Sonnet. Opus 5.5 has been the default on Pro since 22 September 2026 and uses more quota than Sonnet. If Pro stops you often, switch to Sonnet with /model before you pay for Max. /usage shows where you stand.
Does Claude Max work faster than Pro?
Anthropic's pricing page doesn't list speed as a Max benefit. Pro and Max both run Opus and Sonnet. What it does list for Max is higher output limits, early access to advanced features and priority access at high traffic times, on top of more usage per session and Fable models up to 50% of weekly limits. Fast mode is a research preview that runs Opus 5.5 up to 2.5x faster. It's billed through usage credits on Pro and Max alike and doesn't count against plan limits.
What happens when I hit the Claude Code limit?
Claude Code stops and shows which limit you hit and when it resets, for example "You've hit your session limit · resets 3:45pm". You can wait, and recent versions continue the task in the open session after the reset. You can also turn on usage credits with /usage-credits to keep working at standard API rates, use a Claude Console account for API-billed work, or move up a plan. An Opus or Sonnet limit covers only that model family, so switching to another family with /model keeps you working.
Is the API cheaper than a Claude subscription for Claude Code?
Only for light use. Anthropic's docs put API-billed Claude Code at about $13 per developer per active day across enterprise deployments. A plan is cheaper once you use it on more than about 1.5 active days a month for Pro, 7.7 for Max 5x or 15.4 for Max 20x, as long as its limits cover your sessions. At the $30 day that 90% of users stay under, the figures are 0.7, 3.3 and 6.7. The formula is plan price divided by your API cost per active day.
Does Claude Code use my subscription or my API key?
If ANTHROPIC_API_KEY is set in your environment, Claude Code uses the key instead of your plan and bills API usage. Interactive sessions ask you once to approve it, and claude -p uses it whenever it is set. Run unset ANTHROPIC_API_KEY and check /status to go back to your plan.
Does running Claude Code on a server or on Hivra give me more usage?
No. When you sign in with your plan, your limits follow your Anthropic account, wherever the work runs. Anthropic's cloud sessions, Remote Control, a VPS and a Hivra computer all draw on the same allowance. A computer that stays on only affects whether the hours you pay for get used. On a paid Hivra plan the computer stays on and keeps your files, sessions and login. A Claude Code or Codex run you start inside tmux in the computer's Terminal tab keeps going after you close the laptop. On Claude Code, work you send through Telegram, connected in the agent's Telegram tab, runs on the computer, not in your browser. Hivra is independent and is not affiliated with Anthropic or OpenAI.