Claude Pricing in 2026: Every Plan, Real Costs and Hidden Fees

Claude Pro costs $20 a month, or $17 a month on the annual plan, and that annual plan is the right default for most people who write or code with Claude on a normal workday. Max is $100 a month when you need 5x the Pro allowance. A Team standard seat is $25 a month. Free is $0. These figures were verified on Anthropic's pricing page in September 2026.

Claude is Anthropic's assistant for chat, documents, and coding, sold as a flat subscription for people and a per-token API for products you ship, and mixing those two bills is how teams overpay.

Toolradar data: the AI assistants ranking we publish evaluated 1,462 tools, updated September 2026, and Claude's consumer ladder still sits in the same band as ChatGPT rather than above it. Our Claude review covers the product, this page is the bill, including the fees the plan cards leave off, and the product record lives on Claude's Toolradar page.

Plan	Price	Best for	Limit that changes the decision
Free	$0	A real trial, not a demo	Sonnet and Haiku only, 5 projects, no Claude Code
Pro	$20/mo, or $17/mo annual	Daily writing and coding	At least 5x Free per five-hour session, Opus included
Max 5x	$100/mo, monthly only	You already hit Pro limits	5x Pro per session, priority when traffic is high
Max 20x	$200/mo, monthly only	Claude Code most of the day	20x Pro per session
Team standard	$25/mo, or annual at the Pro monthly rate	A shared workspace with SSO	1.25x Pro per person, chats not used for training
Team premium	$125/mo, or $100/mo annual	The two or three heavy users	6.25x Pro per person
Enterprise	$20/seat/mo plus API usage	Compliance, SCIM, audit logs	Annual contract, seat fee is not the whole bill
API	Metered per million tokens	An app you run yourself	Separate from every chat subscription above

Free is a real trial, with a hard ceiling

The free plan is $0, and it is enough to judge the writing, because chat on the web, desktop, and mobile already includes web search, file creation, code execution, memory, connectors, Artifacts, and voice. Projects stop at five, and the models on this tier are Sonnet and Haiku, so anyone who needs Opus, Fable, or Claude Code is already past this plan.

That ceiling is why free stops being free: Anthropic does not publish a fixed message count, and limits reset on a rolling five-hour window, so longer chats on stronger features burn the allowance faster. Pro is at least 5x more usage per session than free. A few questions a week can stay here, while a draft or a bug you finish in Claude will hit the wait before it hits a dollar.

Pro is the plan that should be on the card

Pro is the individual plan, billed monthly or as one annual charge, and the annual invoice is $200 up front. You get more usage than free, Claude Code, Claude Design, Slides, and Docs, Research, and projects without the free cap. Opus joins Sonnet and Haiku, and Claude also runs inside Chrome and Microsoft 365. Fable, the flagship, is available here only through usage credits, not as an open allowance, so heavy use of the flagship is a second bill.

Buy Pro when Claude is in the workday and you are not yet staring at a limit message. Writers, founders, and developers who open it most days will feel the jump from free, because Opus stays available and Claude Code is in the terminal. Claude vs ChatGPT is the product comparison, and the matching ChatGPT bill is in ChatGPT pricing. If you would not notice Claude missing by the afternoon, skip the annual charge.

Max is a capacity plan, not a smarter model

Max is for people who already live in Claude. It includes everything in Pro, higher output limits, early access to advanced features, and priority when traffic is high. The two tiers are monthly only, Anthropic does not sell a standard annual discount on Max, and the published prices are for web checkout, so a mobile app store can charge a different amount than the web price.

Max 5x is five times Pro's usage per five-hour session, and Max 20x is twenty times that Pro allowance. Both tiers also carry a weekly cap across models, reset on a fixed weekday tied to the account rather than the day you subscribed, so a heavy week still ends. Fable is capped at 50% of weekly limits on either Max tier, so the wall moves further out without disappearing.

Skip Max until you can name the Pro limit that is costing you time. Researchers on long tasks and engineers who keep Claude Code open all day are the fit, and everyone else is paying for headroom. Moving from annual Pro to Max can leave a credit when the unused annual value exceeds one Max bill, but only if the billing address stays the same. A Google Play upgrade charges the full new price and turns leftover time into extra days, so the store you use changes the cost of the switch.

Team is an admin product wearing a seat price

Team is for 2 to 150 people. A solo founder cannot open Team to get the business settings, because two seats is the minimum, and past 150 seats Anthropic points you to Enterprise. Standard seats are $25 per member per month on a monthly invoice, or $5 less per member per month when you pay annually. Premium seats are $125 on the monthly plan, or the annual premium rate in the table, and they belong on the two or three heavy users rather than the whole company.

Standard seats include 1.25x Pro's per-session allowance, and premium seats include 6.25x Pro, which is five times a standard seat and the only reason to pay the higher seat. Limits belong to each person, so a heavy user cannot borrow what a teammate did not spend. Weekly caps reset on a fixed clock for the account.

Claude Code is included on every seat, and it still shares that person's pool with chat, so coding and writing draw down the same allowance. Premium seats also get Fable up to 50% of weekly limits. Admins can prepay usage credits so someone can continue at API rates after the included allowance is gone, which is how a flat seat becomes a token bill.

The reason to buy Team is the default on your data, not a lower seat price. Team and Enterprise do not train on your content by default, while Free, Pro, and Max are opt-out, and that default is the point once customer text sits in the chat. You also get SSO, domain verification, central billing, spend controls, shared projects, and enterprise search across Slack, Microsoft 365, and custom connectors.

Domain capture, SCIM, audit logs, and role-based access sit on Enterprise. Anthropic cut these seat prices on January 28, 2026, in the Team pricing change. US list prices exclude tax and can differ by country.

Removing someone does not shrink the invoice, because the seat stays on the plan until an owner reduces the seat allocation. Upgrades from standard to premium are prorated and charged immediately. Anthropic's own billing example is the clean way to budget: ten standard seats on the annual plan cost $2,400 for the year, due up front. Ten standard seats on the monthly plan cost $250 for that month.

Enterprise looks cheap until the tokens arrive

Enterprise is an annual contract: the seat fee from the table, plus usage billed at API rates. Self-serve checkout exists, and so does a sales-assisted quote. That seat buys the Team feature set plus the controls procurement asks for: domain capture, SCIM, audit logs, custom data retention, role-based access, a compliance API, and a HIPAA-ready offering. Sales-assisted deals add invoicing, AWS Marketplace, and a trial, while consumer plans are credit card only, so a team that needs an invoice has to leave the individual plans.

The seat fee is a cover charge, because a team on cheaper models can land under a stack of premium Team seats, while a team that runs Fable all day pays the tokens on top. There is no public usage estimate on the pricing page. If you cannot forecast tokens, ask sales for a quote against a month of real prompts before you sign the annual term.

The API is a second price list

The Claude API does not come with Pro, Max, or Team. You pay per million tokens, input and output, on the Claude Platform, and output costs more than input on every current model, so a long answer is the expensive side of the call. A person chatting in the app should stay on a subscription, and a developer shipping Claude inside a product should be on this meter.

Claude Code's Toolradar page is the coding product, and how it compares with Cursor is a different decision, covered in Claude Code vs Cursor.

Model	Input per million tokens	Output per million tokens	What it is for
Haiku 4.5	$1	$5	Fast, high-volume calls
Sonnet 5	$2	$10	The default for coding and agents
Opus 5.5	$4	Five times the input rate	Daily agentic coding
Fable 5.1	One fifth of the output rate	$50	The flagship, long-running agents

Sonnet 5 was scheduled to rise on September 1, 2026, to $3 per million input tokens and $15 per million output tokens, and that increase was cancelled, so the table is the lasting rate. The Batch API is half price, which is the rate for batch jobs, and Claude 4.6 and later keep a 1 million token window at the normal per-token rate.

Web search is $10 per 1,000 searches, plus the tokens in the results, and a failed search is not billed. Code execution includes 1,550 free hours per organization each month, then $0.05 per hour, with a five-minute minimum, so a short run still consumes that minimum, and Managed Agents add $0.08 per session-hour. US-only inference on Claude 4.6 and later is a 1.1x multiplier, and fast mode doubles Opus 5.5 rates, so either choice costs more than the listed model rate. A cache read on Fable 5.1 is $0.25 per million tokens, and Claude 4.7 and later emit about 30% more tokens for the same text, so an old cost sheet underbills a new model.

What ten people pay against ChatGPT and Gemini

Seat math is where Claude, ChatGPT, and Gemini stop looking interchangeable. The individual rungs match on the monthly step, and the workspace does not. Prices below were checked on each vendor's own plan pages in September 2026.

Buyer	Claude	ChatGPT	Gemini
Light use	Free	Go at $8/mo, then Plus	Free, or AI Plus at $4.99/mo (2x free)
Most workdays	Pro, monthly or annual	Plus, same monthly price as Pro	Google AI Pro at $19.99/mo (4x free)
Heavy daily use	Max 5x	Lower Pro tier, same monthly price as Max 5x	Ultra from $99.99/mo, 5x AI Pro
All day	Max 20x	Higher Pro tier, same monthly price	Ultra at $199.99/mo, 20x AI Pro
10 people, standard, annual	$2,400 for the year	Business standard, the same yearly total	Ten separate AI Pro plans, not one workspace

ChatGPT Plus matches Pro's monthly price and stays monthly, as do Go and Pro, so none of those plans copies the annual individual discount. Its two Pro tiers match Max 5x and Max 20x, and OpenAI paused new sign-ups to the higher one on September 10, 2026, which means a new buyer may not be able to get the all-day ChatGPT tier. Business seats match Team's standard and premium rates, with the same two-seat minimum. Gemini's Plus, Pro, and Ultra prices in the table are personal Google AI plans, not a shared admin console, so ten people do not get one company invoice.

Read the ChatGPT review, ChatGPT on Toolradar, the Gemini review, Gemini vs ChatGPT, and Gemini on Toolradar before you treat the invoices as the same product. For a team that writes and codes, start from the best AI chatbots shortlist. If the job is live-web research, the Perplexity review matters more than a Claude seat count.

Fees that never appear on the plan card

The plan price is the floor. Every plan spends a five-hour session, paid plans add a weekly cap, and chat plus Claude Code share one pool. On Team that pool is per person, so a heavy coder cannot borrow a quiet teammate's leftover usage. Usage credits, billed at standard API rates, are how a flat month stops being flat once the included allowance is gone.

Annual Pro is cheaper per person than an annual standard seat. The gap pays for SSO, central billing, shared projects, and the default that chats are not training data. Five annual Pro plans cost less than five annual standard seats when those controls are not required, so a group of individual users should not open Team for the sake of one invoice. Refunds are generally unavailable outside the UK and EEA 14-day window, or Apple's process for App Store purchases. Cancel at least 24 hours before renewal, or the next period bills.

AIpresso is the short daily brief on which AI prices and products actually moved. Get it here if you would rather not recheck vendor pages every time a seat price changes.

Which plan is worth paying for

For one person, pay annually for Pro, or stay on free. Move to Max only after Pro limits are a weekly event you can describe. Max 20x is for all-day coding, and it has no annual discount.

For a team, put premium seats only on the people who will burn them. Allowances do not pool, so a premium seat for an occasional user is the usual waste. If you need SCIM, audit logs, or a HIPAA-ready setup, price Enterprise only after you model token spend on top of the seat fee.

For a product, ignore the chat plans and price the API. Sonnet 5 should carry ordinary traffic, Haiku is the cheap path, and Fable is a deliberate choice rather than the default. Batch jobs should use the half-price tier. If you are still choosing a coding tool, use the best AI coding assistants guide. Agentic AI is the workflow those meters are built for. People who need prose more than a terminal should read AI writing tools and the ChatGPT alternatives roundup.

How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this breakdown. We checked Anthropic's pricing page, the Team plan article, the Max plan article, and the Claude Platform pricing docs on September 23, 2026. ChatGPT figures come from OpenAI's Plus, Pro, Business, and Go pages. Gemini figures come from Google's US subscriptions page. The 1,462 tool count is from Toolradar's live AI assistants ranking that same month, and no vendor paid for placement. More on how we publish is on the AIpresso about page.

FAQ

How much does Claude cost in 2026?

Free is $0. Pro is $20 a month, or $17 a month on the annual plan with $200 billed up front. Max is $100 a month for 5x Pro usage or $200 a month for 20x, both monthly only. Team standard seats are $25 per person per month, or $20 per person on an annual plan, and premium seats are $125 a month or a lower annual rate. Enterprise is a per-seat fee plus API usage on an annual contract. The API is a separate meter, starting at $1 per million input tokens on Haiku 4.5.

Is Claude free?

Yes. The free plan includes chat, web search, files, Artifacts, voice, memory, and connectors, on Sonnet and Haiku, with up to five projects. It does not include Claude Code, Opus, or Fable. Limits move with how long and how hard each session is, and Pro provides at least 5x the free allowance per five-hour window.

What is the difference between Claude Pro and Max?

Pro is the everyday paid plan, with monthly or annual billing, Claude Code, and Opus. Max adds 5x or 20x Pro's per-session usage, higher output limits, and priority access, and it is monthly only, so the annual discount stops at Pro. Fable on Max is limited to 50% of weekly limits. Max is worth it only when Pro's cap is already in the way of paid work.

How much is Claude Team per seat?

You need at least two members and can go to 150, so one person should stay on an individual plan. Standard seats include 1.25x Pro per session and premium seats include 6.25x Pro, each person spends their own allowance, and annual billing is the cheaper interval on both. Chats are not used to train Claude by default, which is the admin reason to pay for a seat.

Does Claude Code cost extra?

No extra subscription. Claude Code is included on Pro, Max, Team, and Enterprise, and it shares the same usage allowance as chat on that account or seat, so a day of coding can use up the plan before any chat starts. It is not included on the free plan. Heavy coding can be pointed at API credits instead, which bill at the token rates rather than the flat plan.

Is Claude cheaper than ChatGPT?

At the main individual step, no, because ChatGPT Plus matches Claude Pro's monthly price, and ChatGPT's Pro tiers match Max 5x and Max 20x. Claude's annual Pro plan is the one individual discount ChatGPT does not mirror, because Go, Plus, and Pro stay monthly. For ten people on standard annual seats, both Team and ChatGPT Business come out to $2,400 for the year. Gemini AI Pro is $19.99 a month for one person, which is the lower personal price, without Claude's Team admin controls.

What happens when Claude hits a usage limit?

You wait for the five-hour session or the weekly cap to reset, move to a higher plan, or, on a paid plan, turn on usage credits and continue at standard API rates. There is no published message count, because long chats and stronger models spend the allowance faster. On Team, one person hitting their cap does not reduce anyone else's.

Cite this: AIpresso, "Claude Pricing in 2026: Every Plan, Real Costs and Hidden Fees", September 2026.