Claude Pro costs $20 a month billed monthly, or $17 a month on an annual plan, with no per task price. The API charges per token. To compare them, price one task in tokens at list rates and see how many such tasks would add up to your monthly fee. Prices below were fetched on 18 October 2026.
What do the two products charge?
From Anthropic's pricing page, fetched 18 October 2026:
- Claude Pro: $17 per month on an annual plan (billed as $200 up front), or $20 per month billed monthly. Usage limits reset on a rolling five hour session window, with weekly limits on paid plans. Claude Code is included and shares the same usage pool as chat.
- API list prices, from the pricing documentation fetched the same day, are per million tokens (MTok). Claude Sonnet 5.5 is $2 per MTok of input and $10 per MTok of output. Claude Opus 5.5 is $4 input and $20 output. Claude Haiku 5.5 is $0.10 input and $0.50 output for prompts up to 100,000 tokens.
Anthropic does not publish a per task price for Pro, and we make no claim about what either product costs Anthropic to run. This post only does arithmetic on published numbers.
How do you price one task on the API?
Count tokens in and tokens out. A task that reads a lot of code and writes a short answer is input heavy. A task that generates a long document is output heavy.
An illustrative task, with assumptions we chose: the model reads 200,000 tokens of code and context and writes 20,000 tokens of output.
- Sonnet 5.5: 0.2 MTok times $2 is $0.40, plus 0.02 MTok times $10 is $0.20. Total $0.60.
- Opus 5.5: 0.2 times $4 is $0.80, plus 0.02 times $20 is $0.40. Total $1.20.
- Haiku 5.5: 0.2 times $0.10 is $0.02, plus 0.02 times $0.50 is $0.01. Total $0.03.
These are not benchmarks. Your tasks will differ, and agent loops often re-read context many times, which multiplies input tokens. The pricing documentation also lists lower rates for cache hits, which can cut the cost of repeated context. Measure your own token counts before you decide.
Where is the break even?
Divide the monthly fee by your cost per task. At $20 a month and $0.60 per task, 33 such tasks cost the same on the API. Fewer than that and the API is cheaper on paper; more and the subscription is.
Pro is not an unlimited bucket, though. The page states limits on a five hour window and weekly limits. A heavy user can hit them and wait, which has a cost of its own in time. The API has no such window, but spend is uncapped unless you set limits.
What else differs?
Pro is a product for a person using Claude's apps and Claude Code. The API is for building your own software and agents. They have different terms, so check that your use case fits the one you pick, especially for automation.
What to do
- Pick three typical tasks and log input and output tokens for each.
- Price each at the list rate of the model you actually use.
- Divide your monthly fee by that figure to get the break even task count.
- Count how many such tasks you really run per month.
- Check the usage limits against your busiest week, not your average one.
- Re-fetch prices before you commit, because both pages say they change.
Quick check
Is Claude Pro cheaper than the API?
It depends on volume and task size. Under the break even task count the API costs less on paper. Above it, the subscription does, as long as you stay inside its limits.
Can I use my Pro plan to power my own app?
Pro is the consumer and developer app subscription. An app you build calls the API and is billed per token. Read Anthropic's terms for the exact rules.
Sources
- Claude pricing page, fetched 18 October 2026
- Anthropic API pricing documentation, fetched 18 October 2026
Last reviewed: 18 October 2026.
