Claude API pricing and token cost calculator
Updated October 13, 2026
The Claude API charges per million tokens. Opus 5.5 costs $4 for input and $20 for output, Sonnet 5.5 $2 and $10. Cached context costs far less to read back, and batch requests are half price.
- Fable 5.1, input / output
- $10 / $50
- Opus 5.5, input / output
- $4 / $20
- Sonnet 5.5, input / output
- $2 / $10
- Haiku 5.5, input / output
- $0.10 / $0.50
Work out what your tokens cost
Tokens, in millions
New text you send that is not cached
What Claude writes, including its thinking
Context Claude re-reads from the cache
Context saved to the cache
Total cost
$49.82
- Input$0.00 0%
- Output$5.92 12%
- Cache read$31.40 63%
- Cache write$12.50 25%
It starts with a real heavy day of Claude Code on Opus: about 500 requests that would cost $50 at API rates.
API prices per million tokens
| Model | Input | Output | Cache read | Cache write |
|---|---|---|---|---|
| Fable 5.1 | $10 | $50 | $0.25 | $12.50 |
| Opus 5.5 | $4 | $20 | $0.20 | $5 |
| Sonnet 5.5 | $2 | $10 | $0.10 | $2.50 |
| Haiku 5.5 | $0.10 | $0.50 | $0.01 | $0.125 |
Haiku 5.5 prices apply to prompts up to 100,000 tokens. Above that, Haiku costs $0.50 for input and $2.50 for output. Cache write prices are for the standard five-minute cache. Batch requests cost 50% less, fast mode on Opus 5.5 costs 2x, and US-only processing costs 1.1x.
Where the money goes in Claude Code
On that measured day, 63% of the cost was cache reads: Claude re-reading the conversation and code it already had in context, hundreds of times. Output, the code Claude actually wrote, was a small part of the bill.
That is why long sessions get expensive. Starting fresh with /clear between unrelated tasks keeps the context, and the cache reads, small.
API or a plan?
For hands-on coding, a plan costs far less. The measured day above would cost $50 on the API, while Max 20x costs $200 for a whole month. The API makes sense for automation, CI and apps you build. Find the plan that fits you.
Common questions
- How much does Claude Opus cost on the API?
- Opus 5.5 costs $4 per million input tokens and $20 per million output tokens.
- What is a cache read?
- When Claude reuses context it has already seen, such as earlier messages or files, it reads it from the cache. That costs a fraction of normal input.
- Is there a discount for batch requests?
- Yes. Batch requests cost 50% less. They run in the background, so they suit work that does not need an instant answer.