Claude Pricing: Pro, Max, Team and API Costs Explained
Most people should pay $20 a month for Claude Pro, or $17 a month billed yearly. Free is $0. Max is $100 for 5x Pro's usage or $200 for 20x. Team has a per-seat rate set on Anthropic's pricing page. Pro and Max include Claude Code. On the API, Sonnet 5 costs $2 input and $10 output per 1M tokens. Opus 5 is $5 / $25. Haiku 5.5 is $0.10 / $0.50 on prompts up to 100K tokens and $0.50 / $2.50 above that.
Bottom line
Claude's limits run on a clock. Pro gives at least 5x Free's usage per five-hour session, with weekly limits on top. What matters is not your monthly total. It is how often a heavy block of work runs dry before the session resets. If that happens a few times a week, Max fixes it; if it happens once a month, wait it out. Before paying for Max to run Claude Code, price the API. $20 of Sonnet 5 buys 5M input plus 1M output tokens ($10 + $10), and $100 buys five times that. Team is the wrong buy for more personal usage: it is an admin and collaboration plan with a seat minimum.
Current plans and prices
Claude Free
$0/monthLimited access for occasional use. Sonnet 5 is available on every plan and is the default on Free and Pro.
Best for
Trying Claude, short chats and light document work.
Not for
Daily work; the cap will cut sessions short.
Claude Pro
$20/month (US)$17 a month billed yearly. At least 5x Free's usage per five-hour session, plus weekly limits. Includes Claude Code. Regional prices and taxes vary.
Best for
One person writing, researching, coding or analysing most days.
Not for
People who never hit the Free cap, or who run out of Pro's session most days.
Claude Max 5x
$100/monthFive times Pro's usage in one account, for frequent individual users.
Best for
People who hit Pro's session cap several times a week on work that can't wait for a reset.
Not for
Anyone upgrading on a hunch. Count how often Pro stops you for two weeks first.
Claude Max 20x
$200/monthTwenty times Pro's usage, for daily, very high-volume work with Claude.
Best for
People who use Claude as their main production tool all day, including long Claude Code runs.
Not for
Typical professionals. If 5x covers you, 20x is $100 a month of headroom you won't touch.
Claude Team
Per seat; see Anthropic's pricing pageCentral billing, admin and collaboration features. Anthropic sets the per-seat rate and minimum seat count on its pricing page; we don't quote them here.
Best for
Organisations that need managed access and shared administration.
Not for
Solo buyers, or groups smaller than Anthropic's seat minimum.
Claude API — Opus 5
$5 input / $25 output per 1M tokensFlagship model for hard coding and long agent runs; 1M-token context. Costs 2.5x Sonnet 5 on both input and output.
Best for
Tasks where Opus produces an accepted result and Sonnet doesn't.
Not for
Routine volume that Sonnet 5 or Haiku 5.5 completes reliably for less.
Claude API — Sonnet 5
$2 input / $10 output per 1M tokensThe default for coding, agents and professional work; 1M-token context. Batch requests run at roughly 50% of list, and cached input reads at roughly 10%, about $0.20 per 1M.
Best for
Developers building coding, agent and professional-work apps at scale.
Not for
Simple classification or extraction, which Haiku 5.5 does at a twentieth of the price on prompts under 100K tokens.
Claude API — Haiku 5.5
$0.10 input / $0.50 output per 1M tokens (prompts up to 100K tokens)Cheapest current Claude model, released 7 October 2026, at a fiftieth of Opus 5's price; 1M-token context. A prompt over 100K tokens bills the whole request at $0.50 / $2.50, five times more. Haiku 4.5 ($1 / $5) is now a legacy model.
Best for
Extraction, classification, routing and sub-agent steps inside bigger loops.
Not for
Deep reasoning. Move up to Sonnet 5 or Opus 5 when too many answers fail review.
What users report
Chat-heavy users rarely hit Claude's caps. Most complaints and workarounds come from coders hitting the 5-hour and weekly limits.
Paraphrased from public posts, with the date and a link to each original. These are other people's accounts, not our tests, and we have not verified each one.
One user asked Anthropic to fix the 20x Max plan, which they estimated gave only "6-7x" the usage ceiling of Pro.
Limits · Claude Max 20x · Hacker News, Sep 2026
One user said many people buy two Claude Pro subscriptions instead of upgrading to Max, noting "$40 vs $100 is significant for a lot of people."
Value · Claude Pro · Hacker News, Mar 2026
One user reported rarely being rate limited on Claude Pro for questions, data analysis and side projects, and rarely passing 10% session use on Max 20x at work.
Limits · Claude Pro / Max · Hacker News, Apr 2026
One user who tried Claude Pro for a month said it lasted fine for chat but became "useless" once they used Claude Code, so they moved to cheaper models on a $10 plan.
Limits · Claude Pro · Hacker News, Aug 2026
By one user's own measurement, $20 ChatGPT Plus gave about $100 of API-equivalent usage, while Claude Pro gave over $250 (about $300 in their last week, during a temporary +50% boost).
Value · Claude Pro · Hacker News, Jul 2026
One user reported cancelling Claude Pro over unexplained limits they did not see with Codex, even after paying for "extra usage".
Billing · Claude Pro · Hacker News, Apr 2026
Official sources
Frequently asked questions
Is Claude Pro cheaper if I pay yearly?
Yes. Pro is $20 a month billed monthly or $17 a month billed yearly: $204 a year against $240. Pay yearly once a month or two on monthly billing shows you'll keep using it.
Does Claude Max make each message cheaper?
Max 5x doesn't: it is five times Pro's usage for five times Pro's price. Max 20x does: twenty times the usage for ten times the price, so each unit of usage costs half what it does on Pro. That saving only counts if you use the capacity.
What does Claude Team cost?
Anthropic sets a per-seat rate and a minimum seat count on its pricing page, and we link there rather than quote numbers that may move. Buy Team for admin and collaboration. If you want more usage for yourself, Pro ($20) or Max (from $100) is the direct route.
Is Claude Code included in Pro or Max?
Yes. It runs inside your plan's usage limits, so a long coding session counts against the same five-hour allowance. Without a plan it bills at API rates for the model you pick; Sonnet 5 at $2 / $10 is the usual default.
How much does a Sonnet 5 API call cost?
$2 per 1M input tokens and $10 per 1M output. An agent step that reads 20,000 tokens of context and writes 2,000 costs $0.04 plus $0.02, so $0.06. If that context is cached, the read drops to about $0.004 and the step to about $0.024. Batch requests run at roughly half list for work that can wait. Prices as listed on September 11, 2026.
When is Opus 5 worth it over Sonnet 5?
When Sonnet 5's answer would be rejected and Opus 5's would be accepted. Opus 5 is $5 / $25 per 1M, 2.5x Sonnet's rates; Haiku 5.5 at $0.10 / $0.50 takes the cheap steps while prompts stay under 100K tokens. Most agent stacks send the bulk of calls to Sonnet 5 or Haiku 5.5 and escalate to Opus 5 only on failure.
Related decisions
Verified against official vendor pages and our pricing catalogue on September 11, 2026. Regional pricing, taxes, promotions and product entitlements can change. Check the linked vendor page before purchasing.