How to Use the Claude API Cost Calculator
- Pick a model. Opus 5.5 is my default for agent work, Sonnet 5.5 for most production apps and Haiku 5.5 for high volume.
- Enter how many requests you send, per day or per month.
- Enter the tokens in each request: fresh input, output and, if the start of your prompts repeats, the tokens read from and written to the cache.
- Turn on Batch for work nobody is waiting on, US-only if your data has to stay in the United States, and fast mode if you pay for faster Opus output.
- Read the monthly total, then the comparison that runs the same workload on every current Claude model. Click a model to switch to it.
Claude API Prices per Million Tokens (October 2026)
| Model | Input | Output | Cache read | Cache write (5m / 1h) |
|---|---|---|---|---|
| Fable 5.1 | $10 | $50 | $0.25 | $12.50 / $20 |
| Opus 5.5 | $4 | $20 | $0.20 | $5 / $8 |
| Sonnet 5.5 | $2 | $10 | $0.10 | $2.50 / $4 |
| Haiku 5.5, prompts up to 100K | $0.10 | $0.50 | $0.01 | $0.125 / $0.20 |
| Haiku 5.5, prompts over 100K | $0.50 | $2.50 | $0.05 | $0.625 / $1 |
| Haiku 4.5 | $1 | $5 | $0.10 | $1.25 / $2 |
The Batch API halves every price in the table. Output costs five times input on every current model, and thinking tokens are billed as output. A cache read costs 5% of the input price on Opus 5.5 and Sonnet 5.5, 2.5% on Fable 5.1 and 10% on the older models.
Haiku 5.5 has two price levels. A request whose prompt passes 100,000 tokens pays five times more on every line, cache reads and output included. The calculator switches to the higher prices on its own when your prompt crosses that line.
What Moves a Claude API Bill the Most
- Prompt caching. Writing to the 5-minute cache costs 1.25 times input and the 1-hour cache 2 times, while every later read costs a small fraction of input.
- The Batch API. Requests nobody is waiting on cost 50% less, and the discount stacks with caching.
- US-only inference. Setting inference_geo to us adds 10% to every line on Claude 4.6 and later models.
- Fast mode. Opus 5.5 delivers faster output at $8 input and $40 output per million tokens, twice its standard price, and fast mode does not run through the Batch API.
- Effort. Higher effort means more thinking, and thinking is billed as output.
Three Worked Examples
- Support chatbot on Sonnet 5.5: 50,000 requests a month, 2,000 input and 400 output tokens each. That is $400 a month, or $0.008 per request.
- Coding agent session on Opus 5.5: 3,000,000 cached tokens read, 150,000 written to a 5-minute cache, 50,000 fresh input and 10,000 output. One session costs $1.75, against $2.94 on Opus 5. At 440 sessions a month the bill is $770.
- Nightly summaries on Haiku 5.5 with the Batch API: 10,000 documents a month, 8,000 input and 600 output tokens each. The whole month costs $5.50.
For every model, the hidden costs and seven ways to cut the bill, read my guide to Claude API pricing.
Claude API or a Claude Subscription?
Use a subscription when people chat with Claude, and the API when software calls it. Pro costs $20 a month and Max $100 or $200, each with usage limits. The API has no monthly fee and bills every token.
Every plan side by side, from Free to Enterprise, is in my guide to Claude pricing.
If Haiku 5.5 is your model, my guide to Claude Haiku pricing explains the 100,000-token catch with real workloads.
Frequently asked questions
How much does the Claude API cost?
It depends on the model. Per million tokens, Haiku 5.5 costs $0.10 input and $0.50 output on prompts up to 100,000 tokens, Sonnet 5.5 costs $2 and $10, Opus 5.5 costs $4 and $20, and Fable 5.1 costs $10 and $50. Prices from Anthropic's pricing page, checked October 12, 2026.
How do I estimate my tokens?
Send a few real requests and read the usage field in each response, which lists input, output, cache write and cache read tokens. Anthropic also offers a token counting endpoint. Claude 4.7 and later models use a newer tokenizer that produces about 30% more tokens for the same text, so recount when you move from an older model.
Does prompt caching make the Claude API cheaper?
Yes, when the start of your prompts repeats. A cache read costs 5% of input on Opus 5.5 and Sonnet 5.5, 2.5% on Fable 5.1 and 10% on older models. Writing to the 5-minute cache costs 1.25 times input and to the 1-hour cache 2 times.
Is the Claude Batch API cheaper?
Yes. The Batch API costs 50% less on input and output, and the discount combines with prompt caching. It suits work nobody is waiting on, such as nightly summaries or evaluations. Fast mode cannot run through it.
What is the cheapest Claude model?
Haiku 5.5, at $0.10 input and $0.50 output per million tokens on prompts up to 100,000 tokens. Above that line it costs $0.50 and $2.50, which is still half the price of Haiku 4.5.
Sources
Claude Platform Docs, Pricing, read October 12, 2026.