AI APIs

How much does the Claude API cost?

Claude API prices per model, checked in November 2026, how batch processing and prompt caching cut costs, and how to keep a Claude feature from overspending.

The Claude API is billed per million tokens, with output priced above input. On 9 November 2026, Claude Haiku 5.5 cost 0.10 dollars input and 0.50 output for prompts up to 100,000 tokens, Claude Sonnet 5.5 cost 2 and 10 dollars, and Claude Opus 5.5 cost 4 and 20 dollars. Batch processing halves these prices.

Published 9 November 2026 · Last updated 9 November 2026

By Michal Měřínský. Audit methodology.

Prices per model

Anthropic's pricing page on 9 November 2026, per million tokens, input then output:

Claude Fable 5.1: 10 dollars and 50 dollars.
Claude Opus 5.5: 4 dollars and 20 dollars.
Claude Sonnet 5.5: 2 dollars and 10 dollars.
Claude Haiku 5.5: 0.10 dollars and 0.50 dollars for prompts up to 100,000 tokens, 0.50 and 2.50 dollars above that. Older models remain available at their own prices, listed on the same page. Prices change, so check before budgeting.

Ways to pay less

Batch API: 50 percent off input and output for asynchronous requests that can wait.
Prompt caching: a long, repeated prefix, such as instructions or a document, is written to the cache once and read back at a fraction of the input price. Five-minute cache writes cost 1.25 times the input price, one-hour writes 2 times, and cache reads cost 0.1 times by default, with lower rates on some newer models.
Model choice: use Haiku for simple, high-volume tasks and reserve larger models for work that needs them.
Output limits: set max_tokens to what the feature really needs.

Estimating a feature's cost

Multiply average input and output tokens per request by the per-million prices, then by expected requests. A request with 2,000 input tokens and 500 output tokens on Sonnet 5.5 costs about 0.004 plus 0.005 dollars, under a cent. Ten thousand of those a day is roughly 90 dollars a day, which is why limits matter.

Rate limits and tiers

Accounts move through usage tiers as they spend, and each tier has rate limits per model. Hitting them returns 429 errors. Check your tier in the Claude Console and plan for retries with backoff.

Subscriptions are different

Claude Pro, Max, Team and Enterprise are subscriptions for using Claude and Claude Code as a person or team. They do not include API usage for your own app, which is billed separately in the Claude Console.

Set a workspace spend limit in the Console before you launch. It is the backstop when everything else fails.

What a scan can and cannot tell you

WebAuditOS checks the parts of an AI-powered site visible from outside: CORS on your endpoints, HTML error pages with recognizable debug output and third-party script load.

It cannot see your Claude usage, costs or keys.

The cheapest Claude bug to fix is the one that never ships: a public endpoint without a rate limit.

Short answers

Is the Claude API free? No. It is billed per token.

Which Claude model is cheapest? Claude Haiku 5.5, as of 9 November 2026.

Does the Batch API really halve the price? Yes, for input and output, on requests that can be processed asynchronously.

Does Claude Pro include API credits? No. API billing is separate.

In the product

The checks this maps to

Every check named here runs on every scan, on every plan - what a plan changes is how many pages are crawled, how long the history is kept and what you can do with the result. These are the ids the findings carry, so you can match a line in a report to the explanation above.

sec.cors.wildcard - CORS is not wide open
sec.error-disclosure - Error pages do not leak stack traces
sec.third-party-scripts - Third-party script count is controlled

Run a free audit See a full sample report See the whole check library