Claude Pricing 2026: Every Plan, API Rate, and the Fable 5 Credit Catch

Claude pricing runs from free to $200 a month, with API rates billed separately per token. We checked every number on Anthropic's official pricing page on July 10, 2026: Pro costs $20 a month, Max costs $100 or $200 depending on the usage multiplier, and Team seats start at $20. Four plans sound simple. Billing details hide the catches, and this page covers those too.

Two different audiences pay for Claude, and Anthropic prices them separately. Subscriptions cover the chat apps plus Claude Code and Cowork; the API bills per token for anything you build. Mixing up the two mental models is where most budgeting mistakes start.

Free Plan: Full Features, Capped Usage

Anthropic's free plan covers more ground than most rivals give away: chat on web, iOS, Android, and desktop, web search, memory across conversations, file creation with code execution, and connector support through remote MCP. Sonnet 5 handles the requests. You pay nothing.

Limits are the trade. Free usage caps are dynamic - Anthropic adjusts them with traffic and does not publish exact message counts - so a long session can end without warning. Anyone hitting that wall more than twice a week has outgrown the tier.

Claude Pro at $20: The Default Choice

Pro costs $20 a month billed monthly, or $17 a month on the annual plan - $200 billed up front. Beyond a bigger usage allowance, that $20 unlocks the tools that turned Claude into a work platform: Claude Code, Cowork, Design, Science, Research, unlimited projects, and the Microsoft 365 integration.

Most individual buyers should stop reading here. Pro covers daily coding, writing, and analysis for one person comfortably, and the annual option returns $40 a year against monthly billing - one of the few AI subscription discounts that amounts to anything.

Claude Max: $100 or $200 for 5x or 20x Usage

Max comes in two multipliers: $100 a month buys roughly 5x Pro's usage, $200 buys 20x, and both add higher output limits, early access to new features, and priority capacity at peak times. Anthropic never publishes what one unit of "usage" equals in messages or tokens, which makes the multiplier tiers easier to sell than to evaluate.

Heavy Claude Code sessions are what push people up this ladder. If Pro locks you out most days, $100 buys the headroom; the $200 tier exists for people running agents more or less continuously.

Team and Enterprise: Seats From $20

Team pricing uses two seat types that mix freely in one plan: standard at $20 per seat a month billed annually ($25 monthly) with more usage than Pro, and premium at $100 per seat annually ($125 monthly) with 5x that. Plans cover 5 to 150 people and add SSO, central billing, enterprise search, and admin controls.

Enterprise flips the model: $20 per seat plus usage billed at API rates, with admin-set spend limits per user and per organization. Companies with a few heavy users and many light ones tend to come out ahead on that structure.

Claude API Pricing: $1 to $50 per Million Tokens

Developers pay per million tokens (MTok), each model priced separately. Batch processing halves every rate for asynchronous jobs, and prompt caching cuts repeat-context reads to a tenth of the input price or less.

Sonnet 5's $2/$10 rate is introductory - standard pricing of $3/$15 starts September 1, 2026, per the pricing page footnote. Anthropic launched the model at that discount in June, a move we covered when Sonnet 5 became the default.

Platform extras carry their own meters. Web search costs $10 per 1,000 searches on top of tokens, code execution adds $0.05 per container-hour after a free daily allowance of 50 hours, and Managed Agents runtime bills at $0.08 per session-hour. Small numbers one by one - in agent-heavy products they grow faster than the token line does.

Four Billing Catches to Know

Claude pricing changed twice in the first ten days of July alone, so treat every number here as dated the moment Anthropic ships its next billing update. We re-verify this page on a set schedule; the date at the top is real.

Fable 5 is the big one. Anthropic pulled its frontier model out of included plan usage on July 8 - subscribers now get monthly usage credits instead, $10 on Pro and $50 on Max, as we reported when the credit system landed. Exhaust the credit and Fable 5 waits until the next monthly refill; the other models keep working on normal plan limits.

Three smaller ones deserve a look. Annual billing means paying the full $200 up front, not a discounted monthly charge. Sonnet 5's API discount expires August 31, which raises a running Sonnet workload's cost by 50% overnight. And US-only inference, for workloads that must stay stateside, runs 1.1x the standard token rates.

Which Claude Plan Should You Pick

Claude pricing rewards matching the plan to your lockout frequency, not your ambitions. Casual users should stay free. Anyone working with Claude daily belongs on Pro annual at $17 a month, and only sustained Claude Code or agent workloads justify Max - start at $100 and move to $200 only after the 5x tier locks you out too.

Developers building products skip subscriptions entirely and price out the API by model. Haiku 4.5 at $1/$5 covers far more production use cases than most teams expect, and batch mode halves it again.

For the wider picture - how Anthropic's rates stack up against OpenAI, Google, and xAI - see our full guide to AI pricing, and our ChatGPT pricing breakdown covers the closest rival, plan for plan.