Claude pricing in 2026: plans, real limits, and which one is right for you

Claude has one free plan and four paid ones: Pro costs $20 a month ($17 a month if you pay for the year), Max is $100 or $200, Team is $25 per seat ($20 on annual billing), and Enterprise is $20 per seat plus usage at API rates. These are Anthropic’s prices in dollars, before tax; in Spain, VAT is added on top. For most people, Pro is enough. Max only pays off if you code with Claude Code several hours a day or have an agent running nonstop.

Below you’ll find each plan and what it includes today, how the limits really work, API prices per million tokens, a comparison with ChatGPT and Gemini, and a rule of thumb for choosing. All figures were checked on October 16, 2026 against the official pages listed at the end.

Claude’s plans and what they cost today#

Prices in US dollars, before tax, per claude.com/pricing and the Claude Help Center as of October 16, 2026.

Free: for trying it out and the occasional question#

The free plan gives you chat on web, desktop, and mobile, web search, memory, connectors, skills, artifacts, file creation with code execution, and up to 5 projects. It uses the Sonnet and Haiku models; Opus and Fable aren’t included. Neither is Claude Code.

Anthropic doesn’t publish how many messages Free gives you. The only official figure is the comparison: Pro offers “at least 5x more usage per 5-hour session” than Free.

Pro: the default plan#

Pro adds Claude Code (terminal, IDE, and desktop), Opus, Research, unlimited projects, Claude in Chrome, and the Microsoft 365 add-ins. It’s the only individual plan with annual billing: $200 up front, about $17 a month.

The Fable models (Fable 5 and 5.1, the top tier) aren’t part of Pro’s included usage. You can use them, but they’re paid separately with usage credits at API rates.

Max 5x and Max 20x: more usage, not more features#

Max has the same features as Pro with more headroom: 5 or 20 times Pro’s usage per 5-hour session, priority at peak times, and early access to new features. It’s billed monthly only.

Two differences matter:

• Fable is included in the plan: you can spend up to 50% of your weekly limit on it at no extra cost. Max also has its own separate weekly limit for Fable.
• Monthly API credits: $100 on Max 5x and $200 on Max 20x for the Claude Platform. They work for the API, the Agent SDK, and Managed Agents, but not for interactive Claude Code or for extra usage inside Claude. They don’t roll over from month to month, and you need to have been on the plan for 7 days to claim them.

If you run the numbers with the official multipliers, a difference shows up that the table doesn’t make obvious: Max 5x costs five times Pro and gives five times its usage, so usage per dollar is the same. Max 20x costs ten times Pro and gives twenty times its usage: twice as much per dollar as Pro.

Desliza para ver la figura completa

Team: for teams of 2 to 150 people#

Team adds centralized billing, SSO, enterprise search, and admin controls. You can mix standard and premium seats in the same organization.

API credits exist here too: $20 per standard seat and $100 per premium seat, pooled together with a cap of $500 a month. Fable works the same way as on the individual plans: included on premium seats (up to 50% of the weekly limit) and paid with usage credits on standard seats.

Enterprise: the seat gives you access, usage is billed separately#

Self-serve Enterprise costs $20 per seat per month, billed annually, with a minimum of 20 seats. The seat doesn’t include any usage: every token of chat, Claude Code, or Cowork is charged at API rates. In exchange, there are no per-session or weekly limits, and admins set spending caps per person or per organization. It adds SCIM, audit logs, custom data retention, and a HIPAA option.

Education and discounts#

Pro and Max don’t have standing discounts, not even for students; just occasional promotions. What does exist:

• an Education plan for entire universities, with pricing on request;
• reduced pricing on Team and Enterprise for eligible nonprofits.

What Claude costs in euros#

Anthropic doesn’t publish a price list in euros. Its pricing page gives dollar prices “excluding applicable taxes,” and the Pro plan help article explains that the price is shown in local currency where available. It also clarifies that, depending on the region, tax is either included in the displayed price or added at checkout. Tax is calculated from your billing address, which on self-serve plans is your card’s address.

For an individual in Spain, that means VAT (IVA) at the standard 21% rate. As a reference: $20 + 21% = $24.20 a month for Pro, and $121 or $242 for Max. What you end up paying in euros depends on how Anthropic presents it at checkout and on the exchange rate. You’ll see the exact amount at claude.ai/upgrade before you confirm.

In the EU and the UK, you can request a refund within the first 14 days, prorated based on what you’ve used.

How usage limits work#

This is where people get it wrong. Paid plans don’t give you “X messages”; they give you a budget that drains faster or slower depending on what you do. There are two counters:

1. 5-hour session. Resets every five hours.
2. Weekly limit. Covers all models and resets once a week, on a fixed day and time for your account, regardless of when your subscription started. You can see it in Settings > Usage. Max has a second weekly limit just for Fable.

What uses up the most, according to the official help: long conversations, large attachments, Research or web search, the model (Opus and Fable use more than Sonnet), and the effort level. Anthropic reserves the right to set other caps to distribute capacity.

Claude Code shares the same counter as chat, desktop, and mobile. An intense morning in the terminal will show up later in chat. The same goes for the Agent SDK and claude -p: according to the help article updated on October 7, 2026, they still draw from your subscription limits.

When you hit the limit, you have three options:

• Wait for the next reset.
• Turn on usage credits (on Pro, Max, and Team): you keep working and pay for the overage at API rates, with a monthly spending cap that you set.
• Use a “limit reset” if Anthropic has given you one: it sets the session or the week back to zero from Settings > Usage.

Which models each plan includes today#

The current models as of October 2026 are Fable 5.1 (top tier), Opus 5.5, Sonnet 5.5, and Haiku 5.5. According to Anthropic, Mythos 5.1 exists, but access is restricted to verified organizations.

On paid plans, the context window goes up to 1 million tokens on the latest models (Opus 5.5, Sonnet 5.5, Fable 5.1). That’s more than most people need in a chat, but in Claude Code it comes in handy with large repositories.

API prices per million tokens#

The API is pay-as-you-go, with no fixed fee. Input tokens (what you send), output tokens (what the model generates), and cache operations are billed separately.

Prices in dollars per million tokens per the official pricing documentation, checked on October 16, 2026.

What’s worth knowing to avoid surprises:

• Prompt caching. Writing to the cache costs 1.25 times the input rate (5 minutes) or 2 times (1 hour). Reading from it costs very little: 5% of the input rate on Opus 5.5 and Sonnet 5.5, and 2.5% on Fable 5.1. For an agent that keeps resending the same instructions and the same history, caching is the biggest cost-saving lever.
• Batch. Asynchronous requests cost half, and the discount stacks with caching. It’s for anything that isn’t urgent: classifying, summarizing, or translating in bulk.
• Long context with no surcharge. Opus 5.5, Sonnet 5.5, and Fable 5.1 charge the same rate for a 900,000-token request as for a 9,000-token one. Haiku 5.5 is the exception: above 100,000 prompt tokens it moves to the higher rate.
• Opus 5.5 fast mode. It responds up to 2.5 times faster and costs $8 per million input tokens and $40 for output, double the normal rate.
• The tokenizer counts more. According to the pricing documentation, since Claude 4.7 the same text produces roughly 30% more tokens. Keep that in mind when comparing with older prices.
• Other charges. Web search costs $10 per 1,000 searches. Forcing inference to run only in the US multiplies everything by 1.1.

Subscription or API: when each one makes sense#

With a subscription, the price is fixed and the limit is on usage. With the API, you pay for every token, with no ceiling. The comparison depends on how much context you move around, not on how many messages you write.

Here’s a back-of-the-envelope calculation with round numbers and Sonnet 5.5, not counting caching. Imagine 40 messages a day in conversations where each message carries about 15,000 tokens of history and gets an 800-token response:

• Input: 15,000 × $2 / 1,000,000 = $0.03
• Output: 800 × $10 / 1,000,000 = $0.008
• Total: about $0.038 per message, which at 40 a day for 30 days comes to about $46 a month

On Pro you pay a flat $20 as long as you stay within the limits, which Anthropic doesn’t translate into messages. For someone who chats every day, the subscription usually comes out cheaper.

With Claude Code the gap widens dramatically. The official help points out that every turn resends the whole conversation, CLAUDE.md, and the files that have been read. Say 100 turns with 80,000 tokens of context: that’s 8 million input tokens. On Sonnet 5.5 that’s $16 without caching, or $0.80 if everything is read from the cache. On Opus 5.5, $32 or $1.60. That’s why, for coding every day, the subscription wins by a wide margin.

The API makes sense in other cases:

• A product or service for others: a chatbot on your website, a ticket classifier, or a process that calls Claude from your code with nobody at the keyboard.
• Low, irregular volume: a script that runs a few times a month costs pennies.
• Batch jobs with Batch and Haiku 5.5, where a million input tokens costs 5 cents.
• When you need to predict the cost per operation and can’t afford to have a weekly limit stop the process.

If you’re on Max or Team, use up your monthly API credits first ($100 or $200 on Max, up to $500 on Team) before reaching for your wallet.

Which plan to choose based on how you use it#

Desliza para ver la figura completa

The rule of thumb, profile by profile:

• Occasional use. If you ask something a couple of times a week, Free is more than enough. Paying for Pro just to get Opus isn’t worth it if you never even reach the free limit.
• Daily chat. Pro. It’s the jump you’ll notice most: more usage, Opus, unlimited projects, and Claude Code.
• Coding with Claude Code a few hours a day. Start on Pro and watch Settings > Usage for two weeks. If you get cut off mid-morning often, go to Max 5x. Before upgrading, try what the official help recommends: /clear between tasks, Sonnet for routine work, and Opus only for the hard stuff.
• An agent that works all day. That’s our case: at GamzyHost we use Claude Code every day as the foundation of our own assistant, which spends many hours chaining together coding, server, and content tasks. That pattern uses far more than a person typing. With long sessions, subagents, and MCP tools, Pro’s limit quickly falls short. Max 20x is the sensible starting point, and keep a closer eye on the weekly limit than on the 5-hour one.
• Business. From 2 to 150 people, Team, with premium seats only for those who code or use agents heavily. Beyond 150, or if you need auditing, SCIM, or HIPAA, Enterprise. Heads-up: on Enterprise all usage is billed at API rates, so set spending caps from day one.

Signs you’ve outgrown your plan#

• You see the 5-hour limit warning several times a week and lose time waiting.
• You run out of your weekly limit before the reset day. That’s the clear signal: the 5-hour session recovers on its own, the week doesn’t.
• You’ve turned on usage credits and your extra spending for the month is approaching the price difference to the next plan. If you’re paying more than $80 a month in credits on Pro, Max 5x works out cheaper.
• In Claude Code you often see “Usage limit reached · wrapping up.” According to the official help, on Pro that grace period to finish the current step is only granted once a week.
• You avoid Opus or Fable for fear of burning through your limit, even when the task calls for it.

The opposite happens too. If after a month on Max you’ve never gone past 30% of your weekly limit, drop down to Pro and pocket the difference.

Claude vs. ChatGPT and Gemini#

Monthly prices in US dollars, checked on October 16, 2026. ChatGPT’s pricing page (chatgpt.com/pricing) returned an access error, so OpenAI’s figures come from its help center. Google’s come from the US version of gemini.google/subscriptions; we weren’t able to confirm euro prices for Spain on the official page.

The prices are nearly identical: $20 for the standard plan and $100 or $200 for heavy use; Claude and Google even sell them with the same multipliers, 5x and 20x. The difference lies in what each one includes:

• ChatGPT includes Codex in its Pro plans and adds a Pro 500 tier with Ultrafast mode for GPT-6 Astra and GPT-6.1 Sol.
• Google bundles storage (5 TB on AI Pro), YouTube Premium Lite, and Jules, its coding agent.
• Claude bets on Claude Code and the API credits included with Max and Team.

What to do now#

1. If you already pay, open Settings > Usage and check what percentage of your weekly limit you end each week at. That number tells you more than any table.
2. If you don’t pay, use Free for a week. If you hit the limit or miss Opus or Claude Code, move to Pro. Paying for the year only makes sense once you know you’ll stick with it.
3. If you code, install Claude Code and sign in with your subscription, not an API key. You’ll find the steps in the Claude Code guide.
4. If you want to go further and build an agent that works for you, take a look at real-world AI agent examples before choosing between Max and the API.