Claude Pricing: Free, Pro, Max, Team and API Costs, Explained

How much does Claude cost? In the US, Claude has a Free plan at $0, Pro at $20 a month (or $17 a month if you pay $200 for the year), Max at $100 a month for 5x the usage of Pro, and Ultra at $200 a month for 10x Pro's weekly usage with no five-hour limit for verified accounts. For companies, Team seats cost $25 a month ($20 billed annually), premium Team seats $125 ($100 billed annually), and Enterprise is $20 a seat per month billed annually, plus usage at API rates. Developers pay per token on the API, from $0.10 in / $0.50 out per million tokens (Haiku 5.5, on prompts up to 100,000 tokens) up to $10 / $50 (Fable 5.1, the most capable model on the list). Plan prices exclude tax.

And yes, Claude is free to use, with limits that I'll spell out below: the free plan is more generous in some places than you'd expect and stingier in others.

I'm Miguel Ángel, and I wrote this with help from an LLM to verify facts. I checked every price on Anthropic's pricing page and Help Center on November 1, 2026, in US dollars. Small confession: I'm in Spain, and the page showed me euros the first time, because it picks the currency from where your browser connects. If you're outside the US and the numbers don't match mine, that's why; the US version is the one below.

Claude plans and prices at a glance

Two things on that page are easy to miss. First, Ultra's "No 5-hour limit" has a footnote: it's for verified accounts (more on that below). Second, the fine print says prices don't include tax, so your card will show a bit more depending on where you live.

Is Claude free?

Yes. Anyone in a supported country can sign up at claude.ai with an email address and use Claude for $0. The Free plan keeps more than you might expect. According to the comparison table, on Free you get:

• Chat on the web, desktop apps, iPhone and Android.
• Claude Design, Slides and Docs, which used to start at Pro.
• Web search, memory across conversations and voice mode.
• File creation and code execution, so Claude can create documents and run code for you.
• Artifacts, Skills, Connectors to your apps, and desktop extensions.
• Up to 5 Projects.
• The Sonnet and Haiku models, with a context window of "up to 1M" tokens that varies by model.

What Free doesn't include is where the upgrade pitch lives: no Opus and no Fable, no Claude Code, no Research, no Claude in Chrome or Microsoft 365, and no usage credits to keep going once you run out.

The limit that matters on Free

Every Claude plan has usage limits that reset on a rolling five-hour window, and paid plans add a weekly limit on top. Anthropic doesn't publish a fixed number of messages: how far you get depends on how long your chats are, the files you attach and the model you pick. On Free, when you hit the wall, you wait for the reset. On paid plans you can buy usage credits and keep working at API rates.

My take: if you open Claude a few times a week to ask something, draft an email or summarize a PDF, Free is enough and paying would be buying limits you won't touch. The day you catch yourself waiting for the five-hour reset, or wanting Opus for a hard problem, that's your sign.

One more thing about privacy, since it's free and people ask: on Free, Pro, Max and Ultra, whether your chats are used to train models is an opt-out setting. On Team and Enterprise, the page says there's no model training on your content by default.

Claude Pro: what $20 a month buys

Pro is the plan most people should look at first. It's $20 a month, or $200 a year paid up front, which works out to $17 a month. For that you get everything in Free plus:

• More usage per five-hour session than Free (with a weekly limit on top).
• Opus, which Anthropic positions for agentic coding and enterprise work.
• Claude Code, sharing the same usage pool as your chats.
• Research, Projects without Free's cap of 5, and the option to hand off and schedule tasks.
• Claude Science, Claude Dashboards (beta), and Claude in Chrome and Microsoft 365.
• Fable only through usage credits, meaning you pay for it on top at API rates.

The catch Anthropic spells out in its Help Center: Pro doesn't include API usage. If you also want to build something with the Claude API, that's a separate account in the Claude Console with its own bill. Your $20 covers the apps and Claude Code, not your own software. Max, Ultra and Team now include a monthly API credit (more on that below); Pro doesn't.

Is paying annually worth it? If you already know you'll use Claude for a year, paying $200 instead of 12 × $20 saves you $40. If you're not sure, go monthly for a month or two first: Anthropic says payments are generally non-refundable unless its terms or local law say otherwise, and a $200 impulse buy stings more than a $20 one.

When Pro runs out: Max or Ultra

Above Pro there are two plans, both billed monthly only:

• Max: $100 a month, which Anthropic puts at 5x the usage of Pro, both per five-hour session and per week.
• Ultra: $200 a month, which Anthropic puts at 10x Pro's usage per week, with no five-hour limit for verified accounts.

If you remember the old lineup, Ultra is the plan that used to be called Max 20x, at the same price. The "20x" is gone from the page: Anthropic now describes Ultra as 10x Pro's weekly usage, and the five-hour window doesn't apply once your account is verified.

Both plans include everything in Pro, plus Fable for up to half of your weekly limit, early access to advanced Claude features and priority access at high-traffic times.

And both come with monthly API credits: $100 a month on Max and $200 on Ultra, to spend on the Claude Platform (Anthropic's API) on any model. If you were already paying for the API on the side, that changes the math: the plan now covers part of that bill. The small print from the Help Center: you can claim them after seven days on the plan by linking one Claude Console organization, unused credits expire at the end of each billing cycle, and they don't pay for interactive Claude Code sessions or extra usage.

Here's the math I find useful. Measured by the week, Max and Ultra cost the same per unit of usage as monthly Pro, going by Anthropic's own multipliers: Max is 5 times the price for 5 times the usage, and Ultra is 10 times the price for 10 times the weekly usage. So moving up never buys you a bulk discount; what you're paying for is not being stopped. On Max you still have a five-hour window (5x Pro's). On a verified Ultra account you don't, so only the weekly limits are left (one across all models and a separate one for Fable). And none of it makes sense unless you'd actually use it.

When I'd move up:

• You hit Pro's five-hour limit several times a week and you're using Claude for paid work. At that point, the $80 difference to Max is probably worth it.
• You use Claude Code for hours at a time. It draws from the same pool as your chats, so long coding sessions and chatting add up against one limit.
• You want Opus as your default for long sessions rather than something you save for special occasions.
• You're already on Max and still run out, either at the five-hour mark in the middle of a long job or before the week is over. That's the one case for Ultra: twice Max's weekly usage, no five-hour window once your account is verified, for another $100.

If you hit the limit once a week, don't upgrade yet. Anthropic says the model you choose affects how fast you use up your allowance, so save Opus for the hard problems and see how far Pro goes.

Which plan includes Claude Code?

All paid plans do: Pro, Max, Ultra, Team and Enterprise. Free doesn't. Claude Code runs every day on the server where I manage my websites, so the line that matters to me is in Anthropic's own FAQ: Claude Code shares the same usage limits as the rest of your plan. Your terminal sessions and your chats come out of one pool.

For very heavy coding, Anthropic also offers a way out of the subscription limits: pay-as-you-go API credits through a Console account. That turns your coding into a per-token bill (next section), which can be cheaper or much more expensive depending on how much you run it.

Claude Team and Enterprise pricing

Team is for 2 to 150 people, and you can mix standard and premium seats, which is the smart move: give premium seats to the two developers who live in Claude Code and standard seats to everyone else. You can add seats mid-term and pay monthly or annually by card. Team now also includes monthly API credits: $20 per standard seat and $100 per premium seat, pooled into one balance and capped at $500 a month.

A standard Team seat costs a bit more than Pro ($25 vs. $20 monthly), and for that you get more usage than Pro, shared billing, admin controls, SSO and no training on your content by default. For a two-person business that handles client data, I'd take Team over two separate Pro accounts just for that last point.

Enterprise works differently: a low seat price ($20 a month, annual only, with a 20-seat minimum) plus whatever your people use, billed at API rates. That's great if most of your staff uses Claude lightly, and a budget surprise if a few power users go wild. There's a self-serve version and a sales-assisted one with invoicing and tiered incentives on committed spend. Universities have a separate Education plan with no public price.

Claude API pricing per million tokens

The API is for building Claude into your own app, script or automation. You don't pay a monthly fee; you pay for tokens, the chunks of text the model reads (input) and writes (output). Sorry for the jargon; there's no simpler way to say it.

To make that concrete: say your app sends 1,000 requests, each with about 2,000 tokens of input and 500 of output, to Sonnet 5.5. That's 2 million input tokens ($4) plus half a million output tokens ($5): about $9. The same job on Haiku 5.5 comes to about $0.45, and on Opus 5.5 to about $18. Yes, that's a twentieth of the Sonnet bill, and no, you don't get Sonnet's brains for it: Anthropic pitches Haiku 5.5 for quick, repetitive work like summaries, classification and database queries, and as a helper that a bigger model hands smaller jobs to. For complex agentic coding, Anthropic itself still points you to Sonnet 5.5 or Opus 5.5.

A few extras on the same page that change the bill:

• Batch processing cuts prices by 50% if you can wait for results instead of getting them live.
• Prompt caching makes repeated context (a long system prompt, the same document) much cheaper to reread. Anthropic has just halved Sonnet 5.5's cache reads to $0.10 per million tokens, which it says makes Sonnet 5.5 around 20% cheaper on most agentic work.
• Fast mode for Opus 5.5: Anthropic says it's up to 2.5x faster, at 2x the standard price.
• Web search through the API costs $10 per 1,000 searches, on top of tokens.
• Older models are still listed: Sonnet 5 keeps the same $2 / $10 as Sonnet 5.5, older Opus models cost $5 / $25 and Sonnet 4.6 and 4.5 cost $3 / $15, so Opus 5.5 is actually cheaper than its predecessors and there's little reason to stay on the old ones for price. Haiku 4.5 has joined that legacy list at $1 / $5, ten times Haiku 5.5's per-token price on short prompts.

Anthropic's short video on what decides the cost of a task: its case is that picking the cheapest model at the lowest effort isn't always the cheaper choice, and that a more capable model like Fable 5.1 can cost less to complete a task, plus which model and effort to pick for open-ended and for routine work. It's the maker talking about its own models, so check it against your own bill.

3:45

API or subscription: which one do you need?

If a human is typing (you, chatting or coding with Claude Code), get a subscription. It's a flat, predictable price instead of a per-token bill that grows with every long session.

If software is calling Claude (a customer support bot, a script that processes 5,000 documents a night, a feature inside your product), you need the API: what a plan gives you in the apps and Claude Code doesn't cover that. You can have both: Pro or Max for yourself and a Console account for your app. On Max, Ultra or Team, the monthly API credit means a small side project can run on the plan's credit first.

An independent walk-through of Pro, Max and the API from someone who says he runs two Max plans, with a warning about running tools like OpenClaw on a Pro or Max account instead of the API. It's from February 2026, so check today's prices and models against the ones in the video.

5:35

How to pay less for Claude

• Start on Free and only upgrade when the five-hour limit gets in your way.
• Pay Pro annually once you're sure: $200 instead of $240.
• Check the web price first. Anthropic's Help Center says its Max and Ultra prices are for web subscriptions and mobile pricing may vary by app platform.
• Save Opus for the hard problems. Anthropic says the model you pick affects how fast you use your allowance.
• Don't wait for a permanent discount. Anthropic says it has no standing discounts on Pro, Max or Ultra, only occasional limited-time promotions.
• On the API, batch and cache. Half price on anything that can wait is an easy saving.
• Try Haiku 5.5 first for high-volume, simple jobs. Summaries, classification and database queries are what Anthropic designed it for, and on short prompts its per-token price is a tenth of Haiku 4.5's.
• For teams, mix seats instead of buying premium for everyone.

The plan I'd pick, depending on how you use it

If you're weighing Claude against ChatGPT, don't decide on the price tag alone: try both free plans for a week and pay for the one you keep opening. If you code in the terminal, having Claude Code included in Pro is a strong argument for Claude's $20.

Prices and plans are subject to change at Anthropic's discretion, so if something here no longer matches the pricing page, trust the page and tell me through the contact form so I can update this.