Anthropic sells five plan families, and the four that cost money differ less in features than in how much of a shared budget you get. That makes the choice arithmetic rather than aspirational, once you know what each tier actually gates.
Every figure below comes from an Anthropic page. The gaps are labelled as gaps rather than filled in.
The official price table
| Plan | Official price wording | Billing period |
|---|---|---|
| Free | $0 — “Free for everyone” | — |
| Pro | $17 per month with annual subscription discount ($200 billed up front). $20 if billed monthly | Monthly or annual |
| Max | From $100 per month; choose 5x or 20x more usage than Pro | Monthly only |
| Team — Standard seat | $20 per seat / month if billed annually. $25 if billed monthly | Monthly or annual |
| Team — Premium seat | $100 per seat / month if billed annually. $125 if billed monthly | Monthly or annual |
| Enterprise | $20 per seat per month plus usage billed at API rates | Annual only |
Two footnotes travel with the table. The page states that prices “don’t include applicable tax” and that “prices and plans are subject to change at Anthropic’s discretion”. Team covers teams of 2 to 150.
Anthropic’s Max help page gives the two variants their own numbers: Max 5x at $100 per month and Max 20x at $200 per month, both monthly.
Three rules that decide more than price
Rule one: limits are windows, not allowances. The pricing FAQ is explicit — “every plan has usage limits that reset on a rolling five-hour session window, and paid plans add weekly limits on top”. Your activity on web, desktop, mobile and Claude Code “all draws from the same pool”.
Rule two: there is no message count to compare against. The same FAQ states “so there’s no fixed message count”, and names four variables that consume the budget: conversation length and complexity, the features you use, the model, and the effort level. A review that ranks tiers by messages per day is ranking something Anthropic does not publish.
Rule three: when you hit the limit, paid plans have an extra exit. You can wait, upgrade, or turn on usage credits to keep working at API rates. On Free, only the first two exist.
The Fable 5.1 twist
This is the row most comparisons miss. In the “Models and usage” table, Fable reads “No” on Free, “Usage credits” on Pro, and “50% of weekly limits” on Max 5x and Max 20x.
The same model therefore costs differently depending on your tier. A Pro user reaches Claude Fable 5.1 by buying usage credits on top of the subscription. A Max user reaches it by spending up to half the weekly allowance. Opus, the row above, is simply “Yes” on Pro and above.
That asymmetry is the real question behind “should I upgrade to Max”, and the site answers it in a table cell rather than in prose. Fable 5.1 itself is listed on the model pages at $10 per million input tokens and $50 per million output tokens, with a one-million-token context window.
Pro: what $17 or $20 actually buys
The pricing FAQ answers this directly. Pro gives you everything in Free “with more usage and the full set of Claude capabilities”, and lists Claude Code, Claude Cowork, Claude Design, projects, access to more Claude models, and Claude for Microsoft 365.
The relative figure is the one to hold on to: Pro gives “at least 5x more usage per 5-hour session than Free”.
Two things Pro does not include. Priority access at high traffic times is “No” on Pro and appears only on Max. And the Pro help page notes the plan “does not include API usage through the Claude Console” — the API is billed separately, which the API getting-started guide covers.
Max 5x vs Max 20x: how to tell which one you are
Anthropic frames Max for people who “work with Claude throughout the day and want longer, uninterrupted sessions”. The two variants scale that: Max 5x provides five times more usage per session than Pro, Max 20x provides twenty times.
Max adds two things Pro lacks: higher output limits for all tasks, and priority access during high-traffic periods. The Fable row also changes, from “Usage credits” to “50% of weekly limits”.
So the signal is not “I use Claude a lot”. It is whether you are buying usage credits on Pro to get through the week. If you are, and Fable is part of your work, Max 5x is the tier whose terms change. Only that observation, not a projection, should move you to 20x.
Two checkout constraints
The Max help page states both plainly. “The Max plan is currently available as a monthly subscription only”, so there is no cheaper annual rate to wait for. And “these prices are for web subscriptions only”, meaning mobile in-app pricing may differ by platform.
Worked example: three people, three plans
Case one: the professional who opens Claude a few times a day. Short threads, occasional long documents, no terminal work. Pro fits, because the free tier’s gate on Opus and the absence of Claude Code are not constraints here, and Pro’s “at least 5x” session usage covers the peaks.
Case two: the developer who lives in Claude Code and also wants Fable for long-horizon work. Pro is the entry point, but the Fable row says “Usage credits”, so the effective cost is the subscription plus whatever credits the work consumes. Compare that with Max 5x, where Fable is included up to 50% of the weekly limit. If credits are bought most months, Max 5x is the cheaper structure even at a higher list price.
Annual versus monthly, using only published figures:
Monthly Pro — $20 per month over 12 months
Annual Pro — $200 billed up front, expressed as $17 per month
Difference — two dollars a month in Anthropic's own presentation
Not published: the discount percentage behind it, so quote both
figures rather than deriving a percentage.Before you commit, check three published constraints:
1. Max is monthly only, so there is no annual Max rate.
2. Prices exclude tax, which is added according to your region.
3. Enterprise is annual only, and adds usage billed at API rates.Team and Enterprise: when a seat beats a subscription
Team changes two things beyond seat pricing — seats can be mixed, and the plan covers a group with administration attached — and it is described as “for teams of 2 to 150”. Enterprise moves model access up a tier. In the second comparison table, Fable reads “Yes” on Enterprise where it reads “No” on Team, and the context window reads 500k on the default model instead of 200k. Enterprise is annual only, at $20 per seat per month plus usage billed at API rates, with SCIM, audit logs, custom data retention and role-based access — controls the Pro plan does not carry, and which matter if you were weighing Enterprise against simply buying API credits.
What Anthropic does not publish
No fixed message counts on any tier, and no figures for the weekly limits, which are described as existing but never quantified. No conversion rate between usage credits and tokens, so the practical cost of Fable on Pro cannot be computed in advance. No percentage for the annual discount, no price for the Education plan, and no multiple for how much more a Team Standard seat gives than Pro.
Two further gaps: the model matrix does not define what priority access means in practice, and Anthropic states it offers no standard discounts on paid plans and issues no coupons through support.
Frequently asked questions
What is the difference between Pro and Max? Usage per session, higher output limits, priority access during busy periods, and how Fable is metered. Pro lists Fable under “Usage credits”; Max includes it up to 50% of the weekly limit.
Is Max 5x or Max 20x right for me? Look at whether you are already buying usage credits on Pro. That is the published signal that a session allowance, not a subscription, is your limit.
Can I switch or cancel? The FAQ says plans can be changed at any time in Settings, with upgrades immediate and downgrades effective at the end of the billing cycle.
Where does the free tier sit in this? The free-plan breakdown covers what $0 includes, including which models it gates. The topic hub collects the pages behind this one.
Compare the tiers on the two published variables — session allowance and how Fable is metered — and the price gap stops being the deciding factor.