How to Choose a Claude Subscription: Full Breakdown of Pro, Max, and Ultra Pricing and Usage

Title

How to Choose a Claude Subscription: Full Breakdown of Pro, Max, and Ultra Pricing and Usage

Meta Description

How do you pick among Claude's four subscription tiers? This article breaks down the price and benefit differences across Free, Pro, Max, and Ultra, explains the usage rules shared by the five-hour window and weekly limit, distinguishes the three types of quota—plan usage, usage credits, and monthly API credits—and offers tier recommendations by user type along with a team purchasing plan.

Article Content

As of 2026, Claude's individual subscriptions have expanded to four tiers: Free, Pro ($20/month), Max ($100/month), and Ultra ($200/month). The former Max 20X has been renamed Ultra, with existing users migrated automatically and prices unchanged. The real difference among these four tiers isn't the monthly fee figure—it's the usage rules and quota types. Mix them up, and your math will be wrong.

Four Tiers: Pricing and Core Benefits

Free costs nothing. Pro is $20 per month or $200 per year (about $17/month). Max and Ultra are monthly-only, with no annual option. Pro's annual plan saves $40 versus monthly—the only discount across all four tiers.

On benefits: Free lets you chat, search the web, create files and Artifacts, and keep up to 5 projects, but it has no Claude Code and no Opus model. Pro adds Claude Code, Opus, Research, and Claude in Chrome. Max offers 5x usage, plus $100 in API credits and priority access during peak hours. Ultra gives 10x weekly usage, removes the five-hour limit for verified accounts, and includes $200 in credits.

The jump from Free to Pro is the biggest single step; the upper two tiers are mainly about more volume. The official help center explicitly states that Max 20X is now called Ultra, that existing subscribers are migrated automatically at unchanged prices, and that no action is required.

How Usage Is Calculated: Two Windows Sharing One Quota

Claude usage is the sum of two overlapping windows: a rolling five-hour window, which most accounts have, and a weekly limit, which only stacks on paid plans and resets at a fixed time on your account—not from when you start using it.

All four entry points—web, desktop, mobile, and Claude Code—share the same quota. Run code in your terminal for a few hours, and your web chat quota drops accordingly; conversely, chat a lot on the web, and Claude Code will hit the wall sooner too. This isn't a bug—it's by design.

The official documentation gives only multipliers, not absolute numbers: Pro is the baseline (1x per five hours and per week), Max is 5x, and Ultra is 10x per week with no five-hour limit for verified accounts. Why no fixed message counts? Because consumption depends on conversation length, complexity, chosen model, and features—the same question with 200 lines of pasted code versus 20 lines can consume several times as much. The "X messages per five hours" figures floating around online can't be used for planning.

Check your capacity on the Usage page in Settings, where the five-hour and weekly progress bars and credit balance all live. If you don't see a Current Session progress bar, your account has been verified and has no five-hour limit. This verification is determined automatically by the system based on account and payment history—verifying a phone number won't remove the limit.

Don't Confuse the Three Types of Quota

This is the most error-prone part. In Claude's documentation, "quota" refers to at least three different things:

First, plan usage. The five-hour window plus weekly cap, included in the monthly fee at no extra charge. Chat, desktop, mobile, and Claude Code all share this one pool. Free plans have it too, just much smaller.

Second, usage credits. Billed at standard API prices once you exceed your plan, enabled via Enable on the Usage page in Settings. Once enabled, you can set a monthly spending cap, pre-fund the balance, and turn on auto-recharge. This money generally doesn't expire, but in Japan, purchases expire six months after purchase starting September 10, 2026. Think before enabling: this is a credit card with no ceiling.

Third, monthly API credits. Max gets $100/month, Ultra gets $200/month, and Team accumulates per seat. These can only be used for the Claude API, Console Playground, Managed Agents, and the Agent SDK—not for interactive Claude Code—and they don't roll over; they expire each billing cycle. Many people assume the $200 credit always covers Code—the answer is no.

If you know you'll consistently exceed your plan, you can buy discounted usage packs in advance: pay $45 for a $50 pack, $200 for $250, $700 for $1,000—effectively 10%, 20%, and 30% off, provided you enable usage credits first. Individuals can buy up to $2,000 in discounted packs per month; Team up to $3,000. The discount is calculated on face value, not actual usage, so buying more than you use is still waste.

Claude Console is Anthropic's API management backend for developers, a separate system from the Claude.ai chat subscription. In Console you can create an Organization, generate API Keys, and view call records and billing. The monthly API credits included with a subscription only arrive once you "link an organization," because this money is essentially deposited into a Console account. Understanding this is key: chatting via subscription and calling the API are two separate billing pools—the former bills a monthly plan fee, the latter bills per token. Confusing the two is the root of bad math.

Fable Models and Claude Code Billing Details

Fable is a new model with tier-specific treatment. On Pro, Fable 5 and 5.1 don't count against plan usage—they draw from usage credits from the first use, meaning you pay directly. Max and Ultra can run Fable using up to half their weekly quota, with usage credits only kicking in beyond that. The old promotion where Fable counted toward plan usage ended on July 19, 2026. Pro users are most likely to stumble here.

Claude Code doesn't cost extra; Pro, Max, Ultra, and Team all include it, and it shares the same usage pool as other surfaces. But included doesn't mean unlimited—once you've used it up, you still need to enable usage credits or switch to an API account.

After exceeding your plan, there are two paths: the first is enabling usage credits within your subscription, with no need to change login method—chat and Claude Code share one balance, and you can buy discount packs; the second is switching to a Console API account billed by usage, suited to those who need to share API billing with a team or are already doing API development. Both paths charge standard API prices; the difference is billing ownership and management. Running both at once makes it hardest to tell where the money went—pick just one.

There's also an environment variable that can silently change your billing method: if ANTHROPIC_API_KEY is present in your environment, Claude Code authenticates with that key and bills at API prices instead of drawing from your subscription quota. Many people think they're on their subscription, only to discover a Console bill at month's end. To use your subscription, first confirm this variable isn't set—it most often appears when switching computers, changing proxies, or copying environment variables from a tutorial.

ANTHROPIC_API_KEY is the environment variable Claude Code reads, and it takes precedence over a logged-in subscription account. As long as it exists in your shell environment with a value, Claude Code treats it as the authentication credential, and all requests go through Console's API billing channel. To check, run echo $ANTHROPIC_API_KEY in your terminal (on Windows, echo %ANTHROPIC_API_KEY%); any output means the variable is set. Clear it with unset ANTHROPIC_API_KEY (on Windows, set ANTHROPIC_API_KEY=). Note that this variable may be written in .bashrc, .zshrc, or system environment variables—it's most easily introduced when switching computers or copying configuration from tutorials.

Five Types of Users, Five Answers

The first judgment call doesn't depend on budget—it depends on which limit is holding you back.

If you only exceed your plan a few days a month, enabling usage credits with a $30 cap is cheaper than upgrading. If you burn through your quota mid-week every week, upgrade to Max: 5x usage plus $100 in API credits. The most common case is hitting the five-hour limit rather than the weekly cap—that's what Ultra solves.

By user type: for occasional chatting and looking things up, Free is enough, but there's no Claude Code; for daily office work plus light coding, Pro annual at about $17/month, including Code and Opus; for using Code several hours a day, Max; for running long tasks all day, Ultra. This table is editorial judgment—official guidance by user type doesn't exist. If you're torn between two tiers, choose the lower one and use it for a month to see clearly.

Is Pro annual worth it? At $200/year, that's $17/month, saving $40 a year, and you don't have to worry about being locked in. When upgrading from Pro annual to Max monthly, the remaining value of Pro is prorated as credit, and the excess is stored as account balance that automatically pays future bills—so you don't have to wait for the annual term to expire to upgrade. But if you're not sure you'll use it for a full year, start monthly.

Comparisons with similar products: at the $20 tier, Claude Pro, ChatGPT Plus, and GitHub Copilot Pro all cost the same. The lowest entry point is GitHub Copilot Pro at $10, and Cursor starts at $8 with subscription pricing, though using third-party models requires the $30 tier. Claude's differentiator is coverage—one subscription manages chat, documents, Claude Code, and the browser extension at once.

How Teams and Enterprises Should Buy

Team has two seat types: Standard at $20 per person per month annually or $25 monthly, and Premium at $100 annually or $125 monthly, aimed at teams of 2 to 150, and the two can be mixed. Premium seats get 5x the usage of Standard, with Fable rules matching Max. Standard seats follow Pro's Fable rules. The whole team is billed together, with SSO and an admin backend, and team content isn't used to train models by default.

Beyond 150 people or if you need SCIM and audit logs, you'll need Enterprise. Enterprise is priced per seat at $20 per seat per month, with a minimum of 20 seats annually, and usage billed separately at API prices. The 20-seat minimum is a hard threshold—a 19-person company still has to buy 20 seats. If you're just a dozen or so people using it together, Team is enough.

The included API credits form a pool capped at $500. The annual seat fees exactly match the included API credits, and anything beyond $500 isn't granted. So the key to team configuration is seat mixing—not everyone needs 5x usage.

SSO (single sign-on) means team members log in with the company's unified identity system (such as Okta or Azure AD) rather than managing individual Claude account passwords, letting admins centrally grant and revoke access. SCIM is the companion automated user-provisioning protocol: when employees join or leave, the system automatically creates or deactivates accounts in Claude with no manual work. These two are the key dividing line between Team and Enterprise: Team includes SSO but not SCIM or audit logs, and you only need Enterprise when you exceed 150 people or require compliance auditing. For a small team of a dozen or so, Team's SSO is already enough, and there's no need to pay Enterprise's seat minimum just for SCIM.

Seven Steps from Sign-up to Setting a Cap

Step one: sign up at Claude.ai. Anthropic only provides service in supported regions; regions not on the list can't subscribe normally. Step two: choose a plan in Settings under Billing. Upgrades take effect immediately, with unused time on the old plan prorated; downgrades wait until the current billing cycle ends.

Step three: install Claude Code and log in with your subscription account. If you previously logged in with a Console API account, run /login in Claude Code to switch back to your subscription account—this step is critical, since the login method directly determines whether you draw from subscription quota or get billed via API.

Step four, and the easiest to miss, is checking environment variables. As long as ANTHROPIC_API_KEY has a value, Claude Code bills via API and never touches your subscription quota. To use your subscription, unset it first.

Step five: check your five-hour and weekly progress on the Usage page in Settings, and glance at your usage credit balance. Step six: enable usage credits and set a monthly cap—without a cap, consumption can spike fast when agents run long tasks and retry repeatedly. Set it low the first week, then raise it once you understand your rhythm.

Step seven is the monthly API credits available only on Max, Ultra, and Team: after 7 days of subscription, find the API Credits section on the Billing page in Settings and click Link Organization to bind a Claude Console organization. These credits can only call the API and run the Agent SDK—they can't be used in interactive Claude Code—but not claiming them means leaving free money on the table. You only do it once; after that it arrives automatically each month.

Risks and Quick Q&A

Four money-related pitfalls: included API credits don't roll over—use them or lose them that month; usage credits can be set to unlimited, so always set a cap manually; payments are generally non-refundable, and cancellations must be made at least 24 hours before the renewal date; subscribing through the App Store or Google Play may cost a different price than the web.

One more hard limit: Anthropic's supported regions list doesn't include mainland China, Hong Kong, or Macau; Taiwan is on the list. Companies controlled by individuals or entities in unsupported regions aren't covered even if their employees are in supported regions. There are many workarounds online, but accounts and payments can both be revoked, and there's no technical fix for this.

Common questions: Pro is $20 monthly or $200 annually (about $17/month); Max 20X has been renamed Ultra, with existing users migrated automatically at unchanged prices; Claude Code is included in all four paid plans at no extra cost; included API credits can't be used in interactive Claude Code and only call the API—the last one is the most common mistake.

The order for choosing a tier is: first ask what's holding you back, then ask how much you want to spend. Guessing usage from a budget almost always goes wrong.

One-line Summary

Claude's four subscription tiers really differ in usage rules and quota types—pick a tier by identifying which limit is holding you back.

Paragraph Summary

In 2026, Claude's individual subscriptions come in four tiers: Free, Pro ($20/month), Max ($100/month), and Ultra ($200/month), with the former Max 20X automatically renamed Ultra at unchanged prices. The core differences aren't the monthly fees but the usage rules: web, desktop, mobile, and Claude Code share one quota, formed by a rolling five-hour window stacked with a weekly limit. The article focuses on distinguishing three easily confused "quotas"—plan usage, enable-able usage credits, and included monthly API credits—the last of which can't be used in interactive Claude Code and doesn't roll over. It also warns that the ANTHROPIC_API_KEY environment variable can silently switch billing to the API channel, and offers tier recommendations by user type plus team purchasing essentials.