Subscription vs API

Is a subscription cheaper than paying per token?

A plan costs a flat price a month. The API charges for every run, and a run costs more than its token count suggests because each step re-sends the whole context. This page turns a plan price into the number of agent runs it has to beat.

Claude Pro, Max and Ultra prices verified · No signup · Nothing leaves your browser

Compare a plan with the API

Pick a plan, set how many runs you do a day, and describe one run. The workload fields match the main calculator, and the URL keeps your settings so you can share them.

Plan

Editing this switches to Another plan. Taxes are not included.

One run

Standard rates. The batch and cache-write options on the calculator are not carried over.

Break-even

…
···
Runs per day to break even
···
API cost per run
···
API cost per month
···
Plan per month

Break-even by model

ModelPer runBreak-even runs a dayAPI cost at your volume
Cost per run uses the standard rates from the main calculator, with no batch discount and no cache-write charge. A model whose last request is larger than its context window is greyed out. Long-context price tiers are not modelled here; the calculator flags them.

What this cannot tell you

Whether the plan lets you run that many runs. Anthropic states limits only relative to Pro: Max is 5x Pro's usage every five hours and every week, and Ultra is 10x Pro's weekly usage with no five-hour limit for verified accounts. It publishes no token or message count, and the limits are shared between chat and Claude Code. Compare the break-even above with your own usage meter.
  • A plan also includes chat, projects and other features that have no API price.
  • Max and Ultra include $100 and $200 a month of API credit, for API runs only. It expires each billing cycle, so for API work your total with the plan is the larger of the plan price and your API spend.
  • The break-even assumes every run completes the task and prices retries as set above.

Questions

How many agent runs does a Claude Max or Ultra plan have to beat?
Divide the plan price by the cost of one run at API rates. On the default workload (12 steps, Claude Sonnet 5.5, 10% retries, no caching) a run costs $0.351, so Max at $100 a month equals 285 runs a month, about 9.5 a day, and Ultra at $200 equals 570 runs, about 19 a day. With a 90% cache hit rate a run costs $0.096 and the same $100 buys about 1,040 runs, or 35 a day. If your plan lets you run more than that, the plan is the cheaper route.
Why can't this tell me whether my plan covers my usage?
Anthropic states plan limits only relative to Pro. Max is 5 times Pro's usage every five hours and every week, and Ultra is 10 times Pro's weekly usage with no five-hour limit for verified accounts. Pro's own usage has no published number, there is no token or message count, and the limits are shared between chat and Claude Code. With no published quota there is nothing to compare against. The page gives the number of runs your plan price buys at API rates and leaves the comparison with your own usage meter to you.
Does the API credit that comes with Max and Ultra change the answer?
Max includes $100 a month of API credit and Ultra $200. It works for the Messages API, the Batch API, Managed Agents and the Agent SDK, but not for interactive Claude Code or the Claude apps. It expires at the end of each billing cycle and does not roll over. For work you run through the API it only refunds the plan price: your total is the larger of the plan price and your API spend, so the plan cannot be cheaper than the API for that work. The plan wins only where its Claude Code or chat allowance replaces API spend.
Which plans are listed?
Claude Pro ($20 a month, or $17 a month billed annually), Max ($100) and Ultra ($200), with prices read from Anthropic's pricing page on 31 October 2026. Ultra was called Max 20x before Anthropic renamed it. ChatGPT's pricing page shows prices in the visitor's local currency, and no US dollar figure could be confirmed from a primary page, so other plans go under Another plan, where you type the price. Taxes are not included.