Claude Pricing 2026

Claude pricing in 2026: the four individual plans

Anthropic sells Claude on two axes at once, which is where most pricing pages go wrong. There is a ladder of individual plans (Free, Pro, Max 5x, Max 20x) and a separate ladder of seat-based plans (Team, Enterprise). Max is not one product at one price. It is two tiers that differ by a factor of two in both price and capacity.

Plan	Price	Billing	Usage capacity	Context window
Free	$0 (verified 2026-08-18)	N/A	Limited	200k
Pro	$20/mo, or $200/yr (verified 2026-08-18)	Monthly or annual	Standard	200k
Max 5x	$100/mo (verified 2026-08-18)	Monthly	5x Pro capacity per session	200k
Max 20x	$200/mo (verified 2026-08-18)	Monthly	20x Pro capacity per session	200k
Team (standard seat)	$25/seat/mo, or $20/seat/mo billed annually (verified 2026-08-18)	Monthly or annual	Team pooling, 2 to 150 people	200k
Team (premium seat)	$125/seat/mo, or $100/seat/mo billed annually (verified 2026-08-18)	Monthly or annual	Higher pooled capacity	200k
Enterprise (self-serve)	$20/seat/mo plus usage at API rates (verified 2026-08-18)	Monthly or annual	Consumption-based on top of seats	500k on the default model

Prices read from claude.com/pricing and from Anthropic's "Choose a Claude plan" help article, which is itself dated May 19, 2026. Anthropic states that prices shown exclude applicable tax and are subject to change.

The Max mistake worth naming

Anthropic's own plan card says "From $100" and then, in smaller type, "Choose 5x or 20x more usage than Pro." A large number of pages have collapsed that into a single "Claude Max, $200/mo, 5x limits" line, which is two different tiers welded into one wrong claim. The correct reading: $100/mo buys 5x Pro capacity per session, $200/mo buys 20x. If you are quoting Claude Max at $200 and describing 5x limits, you are quoting the price of one tier and the capacity of the other.

What Anthropic does not publish: a message cap

There is no "Claude Pro gives you N messages" number, and any page that supplies one is inventing it. Anthropic's stated model is a rolling five-hour session window, with weekly limits layered on top for paid plans, and the help centre says plainly that how much you can do depends on conversation length and complexity, the model you pick, and the features you use. So the honest answer to "how many messages do I get" is that Anthropic does not publish a fixed message count, and we are not going to guess one for it.

Two knock-on facts that follow from the same design:

• Hitting a limit on a paid plan no longer hard-stops you. Anthropic sells usage credits that let Pro, Max 5x and Max 20x subscribers continue at standard API rates after the included allowance is spent.
• Because limits are session-shaped rather than message-shaped, a single long document analysis can consume more of your allowance than a day of short questions. Capacity per session is what the 5x and 20x multipliers scale.

The context window number, corrected

Claude's "Compare features across plans" table lists a 200k context window on Free, Pro, Max 5x, Max 20x and Team alike. The only larger figure on the page is 500k, on the default model, for the two Enterprise plans. Paying for Max does not buy you a longer context window. It buys capacity, higher output limits, early feature access, and priority at high traffic times. If you specifically need long-context work, the lever is the Enterprise plan or the API, not the Max upgrade.

Claude API pricing

The API is billed per million tokens and is entirely separate from any subscription. Current-generation rates, read from the same page on 2026-08-18:

Model	Input, per 1M tokens	Output, per 1M tokens	Cache write / read
Fable 5	$10.00	$50.00	$12.50 / $1.00
Opus 5	$5.00	$25.00	$6.25 / $0.50
Sonnet 5	$2.00	$10.00	$2.50 / $0.20
Haiku 4.5	$1.00	$5.00	$1.25 / $0.10

Batch processing takes 50% off. Web search as a tool is $10 per 1,000 searches. Code execution gives 50 free container-hours daily per organisation and then $0.05 per hour per container. US-only inference is available at 1.1x the standard input and output rates. Note that the older $3.00 and $15.00 per million figures still circulating for Claude are Sonnet 4.6 rates, which Anthropic now lists under legacy models.

When the Claude free tier is enough

It genuinely is enough for a lot of people, and a page selling a subscription should say so before it sells anything. The free plan is not a crippled demo: it includes chat across web, iOS, Android and desktop, code generation and data visualisation, web search, memory across conversations, file creation and code execution, connectors through remote MCP, and extended thinking. The same 200k context window applies.

Stay on Free if your pattern looks like this:

• You use Claude in bursts a few times a week rather than continuously through a working day.
• Your sessions are short. The limit is capacity within a five-hour window, so occasional use rarely reaches it.
• You do not need Claude Code, Claude Cowork, unlimited projects, or Research, which are the concrete things Pro adds.
• You have a second free assistant to fall back on when you do hit a limit. Meta AI has no paid tier at all, and DeepSeek's app is free end to end.

The signal that Free has stopped being enough is not frustration with quality. It is being interrupted mid-task. If you are hitting the session ceiling more than once or twice a week, Pro at $20 is the cheapest fix available, and it is a smaller step than most people assume, because the model roster and context window barely move. What moves is how long you can keep going.

Claude Pro versus stacking subscriptions

Claude is rarely anybody's only assistant, and that is the actual pricing problem. The commonest second subscription is Gemini, because it is the one wired into Gmail, Docs and Search. Priced at the two numbers verified on 2026-08-18, that pair costs:

Claude Pro $20.00 + Google AI Pro $19.99 = $39.99/mo. Against Perspective AI's Starter plan at $14.99/mo, the difference is $25.00 a month, or $300.00 a year, for both model families and the rest of the roster.

Deliberately absent from that sum: ChatGPT, which is the third subscription most of these stacks carry. OpenAI's US prices could not be re-verified on 2026-08-18 (the pricing page refused automated requests and served non-US currency to a browser), and this page does not quote figures it has not checked. Our ChatGPT pricing breakdown carries that vendor's tiers; add its Plus price to the $39.99 above for the three-subscription version of the same arithmetic.

Perspective AI's Starter plan is $14.99/mo with 200 credits and Pro is $49.99/mo with 700 credits, across a roster of 60+ models with no free tier. Those product figures were last verified internally on 2026-08-07, three weeks before this page was written, so treat them as the older half of this comparison and check the app before quoting them.

The honest caveat: a bundle is not a substitute for Max. If you are on Max 20x at $200/mo you are buying sustained capacity on one model family, and a credit-metered multi-model subscription is a different product solving a different problem. The bundle argument only holds against the $20 tier of two or three vendors at once, which is where most people actually are. Run your own numbers in the subscription cost calculator, or see how the whole field lines up in the master AI plan comparison.

Which Claude plan to choose

• Free, $0: occasional use, short sessions, no need for Claude Code or Research.
• Pro, $20/mo: daily use that keeps hitting the free ceiling. Annual billing at $200/yr saves $40 against twelve monthly payments.
• Max 5x, $100/mo: you hit Pro's session ceiling repeatedly, most days. This is the first tier that is genuinely about capacity rather than features.
• Max 20x, $200/mo: Claude is your primary working surface for most of the day, usually alongside Claude Code.
• Team, $25/seat/mo: two or more people, central billing, and pooled capacity. Annual billing drops it to $20/seat/mo.

For the field outside Anthropic, our breakdown of Claude alternatives ranks the substitutes by price and model access, and AI usage limits compared puts Claude's session model next to the per-message caps other vendors publish.