How to use the Claude API cost calculator
- Pick the model you plan to call.
- Enter the average input and output tokens of one request and how many requests you send per day. The
usageobject of an API response shows real counts. - If you reuse a long prefix such as a system prompt, tool definitions or a document, enter its share of the input and how often it is a cache hit.
- Tick the Batch API for work that can wait. Copy the breakdown or share a link to the same inputs.
How the plan picker estimates
Claude Code sends your conversation with every request. Most of it is read from the prompt cache; new prompts, file reads and tool results are written to it; replies and thinking are billed as output. The picker multiplies one request of each working style by your hours, days, people and model mix, at the prices in the table below. The token counts per style are Claudekit’s assumptions, not Anthropic figures:
| Style | Requests per hour | Cached context | New input | Output |
|---|---|---|---|---|
| Light: questions and small edits | 15 | 25,000 | 2,000 | 800 |
| Typical: feature work with tool calls | 40 | 50,000 | 3,000 | 1,200 |
| Heavy agentic: long sessions, subagents, parallel work | 90 | 90,000 | 4,000 | 2,000 |
For a reality check, Anthropic’s Claude Code costs page
(checked 2026-10-25) reports an average of around $13 per developer per active day across enterprise
deployments, and under $30 per active day for 90% of users. Your own /usage Session block shows the real
figure for a day of your work. Plan prices come from claude.com/pricing
and the Claude Help Center; the rules behind them are in
Claude usage limits explained.
Claude API prices per million tokens
List prices in USD per million input or output tokens (MTok) from the official Claude API pricing page, checked Oct 24, 2026. Batch API requests cost 50% less.
| Model | Input | 5m cache write | 1h cache write | Cache read | Output |
|---|---|---|---|---|---|
| Claude Fable 5.1 | $10 | $12.5 | $20 | $0.25 | $50 |
| Claude Mythos 5.1 (limited access) | $10 | $12.5 | $20 | $0.25 | $50 |
| Claude Fable 5 | $10 | $12.5 | $20 | $1 | $50 |
| Claude Mythos 5 (limited access) | $10 | $12.5 | $20 | $1 | $50 |
| Claude Opus 5.5 | $4 | $5 | $8 | $0.2 | $20 |
| Claude Opus 5 | $5 | $6.25 | $10 | $0.5 | $25 |
| Claude Opus 4.8 | $5 | $6.25 | $10 | $0.5 | $25 |
| Claude Opus 4.7 | $5 | $6.25 | $10 | $0.5 | $25 |
| Claude Opus 4.6 | $5 | $6.25 | $10 | $0.5 | $25 |
| Claude Opus 4.5 | $5 | $6.25 | $10 | $0.5 | $25 |
| Claude Sonnet 5.5 | $2 | $2.5 | $4 | $0.1 | $10 |
| Claude Sonnet 5 | $2 | $2.5 | $4 | $0.2 | $10 |
| Claude Sonnet 4.6 | $3 | $3.75 | $6 | $0.3 | $15 |
| Claude Sonnet 4.5 (deprecated) | $3 | $3.75 | $6 | $0.3 | $15 |
| Claude Haiku 5.5 (prompts up to 100,000 tokens) | $0.1 | $0.125 | $0.2 | $0.01 | $0.5 |
| Claude Haiku 5.5 (prompts over 100,000 tokens) | $0.5 | $0.625 | $1 | $0.05 | $2.5 |
| Claude Haiku 4.5 | $1 | $1.25 | $2 | $0.1 | $5 |
Limited access means the model is offered only to organizations verified through an Anthropic program. For worked examples and tool fees, read the Claude API pricing guide.
Frequently asked questions
How does the calculator work out the cost?
It multiplies the tokens in each part of a request by that model’s list price per million tokens: uncached input, cache writes, cache reads and output. The daily figure multiplies by your requests per day and the monthly figure by 30 days.
How does prompt caching change the cost?
The cached share of your prompt is written to the cache on a miss and read from it on a hit. A 5-minute write costs 1.25 times the input price and a 1-hour write 2 times. A read costs 0.1 times the input price on most models, 0.05 times on Claude Opus 5.5 and Claude Sonnet 5.5, and 0.025 times on Claude Fable 5.1 and Claude Mythos 5.1.
What does the Batch API save?
Batch requests get a 50 percent discount on input and output tokens, and the prompt caching multipliers stack with it. Results arrive asynchronously, so batch suits evaluations, backfills and other work that does not need an instant reply.
Why could my real bill be different?
Token counts are estimates: Claude 4.7 and later models use a tokenizer that produces about 30 percent more tokens for the same text. Tool definitions, server tools such as web search, US-only data residency (1.1x on Claude 4.6 and later), fast mode and taxes are not included.
Can the plan picker tell me how many hours Pro or Max gives me?
No. Anthropic describes plan limits as a five-hour session window and a weekly limit, and says Max 5x and Max 20x give five and twenty times Pro’s per-session usage, but it does not publish either limit in tokens or hours. The picker only compares money: what your described usage would cost on the API against each plan’s price. Use /usage or the status line to see how close you are to your real limits.
Where do the prices come from?
From Anthropic’s official Claude API pricing page, checked on Oct 24, 2026. Retired models are left out. Prices change, so confirm on the official page before you budget.