Claude API cost calculator
Enter your monthly usage and cache hit rate to see what the same Claude workload costs through the Anthropic API, OpenRouter and Wokey. Wokey prices come from the live pricing catalog, the same data billing uses.
Why it costs less and is still the official model
- Official upstream: Claude requests go to Anthropic’s official API. The model a page names is the model you call; it is never swapped for another.
- The saving is on the supply side: Claude supply comes from authorized official subscription quota, which costs less per token than the API list price. The discount applies proportionally to every meter, so the share you save barely depends on your cache hit rate.
- Verify it yourself: Turn on Official verification for a key, and raw passthrough responses carry a proof signed in an AWS Nitro Enclave. The open-source verifier checks the upstream host and response bytes on your machine.
What response proofs do and do not show · When Wokey is the wrong choice
Monthly cost: Claude Code, daily use
A few hours of coding each workday
| Model | Anthropic API | OpenRouter | Wokey | Saving vs official |
|---|---|---|---|---|
| Claude Sonnet 5 | $60.50 | $63.83 | $10.89 | 82% |
| Claude Opus 5.5 | $103.00 | $108.67 | $15.45 | 85% |
| Claude Opus 5 | $151.25 | $159.57 | $22.69 | 85% |
| Claude Haiku 4.5 | $30.25 | $31.91 | $6.05 | 80% |
| Claude Fable 5.1 | $235.00 | $247.93 | $62.28 | 74% |
| Claude Fable 5 | $302.50 | $319.14 | $80.16 | 74% |
| Claude Opus 4.8 | $151.25 | $159.57 | $22.69 | 85% |
| Claude Opus 4.7 | $151.25 | $159.57 | $22.69 | 85% |
| Claude Opus 4.6 | $151.25 | $159.57 | $22.69 | 85% |
| Claude Sonnet 4.6 | $90.75 | $95.74 | $16.34 | 82% |
| Claude Sonnet 4.5 | $90.75 | $95.74 | $16.34 | 82% |
- Anthropic API: Standard global-endpoint price
- OpenRouter: anthropic endpoint at the official token price, plus the 5.5% card fee on credits
- Wokey: No fee for USDT or wallet top-ups; Alipay 3%
Claude Sonnet 5 · Usage preset
| Usage preset | Tokens | Anthropic API | OpenRouter | Wokey |
|---|---|---|---|---|
| Claude Code, daily use | 100M input (90% cache reads, 5% writes), 2M output | $60.50 | $63.83 | $10.89 |
| Claude Code, light use | 25M input (90% cache reads, 5% writes), 0.5M output | $15.13 | $15.96 | $2.72 |
| Claude Code, heavy use | 400M input (90% cache reads, 5% writes), 8M output | $242.00 | $255.31 | $43.56 |
| API app, no caching | 20M input (0% cache reads, 0% writes), 5M output | $90.00 | $94.95 | $16.20 |
Compared with Claude Pro and Max
Subscriptions are a flat monthly fee with no published token allowance: usage is metered in rolling 5-hour windows, paid plans add weekly caps, and Claude Code shares the same allowance as the web app. So a plan has no per-token price; the fair comparison is its monthly fee against your API estimate.
- Claude Pro: $20/month
- Claude Max: from $100/month
Claude model rates
| Model | Uncached input | Cache writes (5m) | Cache writes (1h) | Cache reads | Output |
|---|---|---|---|---|---|
| Claude Sonnet 5 | $0.36 / $2 | $0.45 / $2.5 | $0.72 / $4 | $0.036 / $0.2 | $1.8 / $10 |
| Claude Opus 5.5 | $0.6 / $4 | $0.75 / $5 | $1.2 / $8 | $0.03 / $0.2 | $3 / $20 |
| Claude Opus 5 | $0.75 / $5 | $0.9375 / $6.25 | $1.5 / $10 | $0.075 / $0.5 | $3.75 / $25 |
| Claude Haiku 4.5 | $0.2 / $1 | $0.25 / $1.25 | $0.4 / $2 | $0.02 / $0.1 | $1 / $5 |
| Claude Fable 5.1 | $2.65 / $10 | $3.3125 / $12.5 | $5.3 / $20 | $0.0663 / $0.25 | $13.25 / $50 |
| Claude Fable 5 | $2.65 / $10 | $3.3125 / $12.5 | $5.3 / $20 | $0.265 / $1 | $13.25 / $50 |
| Claude Opus 4.8 | $0.75 / $5 | $0.9375 / $6.25 | $1.5 / $10 | $0.075 / $0.5 | $3.75 / $25 |
| Claude Opus 4.7 | $0.75 / $5 | $0.9375 / $6.25 | $1.5 / $10 | $0.075 / $0.5 | $3.75 / $25 |
| Claude Opus 4.6 | $0.75 / $5 | $0.9375 / $6.25 | $1.5 / $10 | $0.075 / $0.5 | $3.75 / $25 |
| Claude Sonnet 4.6 | $0.54 / $3 | $0.675 / $3.75 | $1.08 / $6 | $0.054 / $0.3 | $2.7 / $15 |
| Claude Sonnet 4.5 | $0.54 / $3 | $0.675 / $3.75 | $1.08 / $6 | $0.054 / $0.3 | $2.7 / $15 |
How it is calculated
- Cache reads are billed at the cache-read rate, cache writes at the write rate for the chosen duration (5 minutes or 1 hour), the rest of the input at the uncached rate, and output at the output rate.
- Claude Code resends a long context every turn, so most input is cache reads. The presets are assumptions for the estimate, not measured bills.
- OpenRouter’s anthropic endpoint charges the official token price; its fee is on buying credits: 5.5% by card (minimum $0.80 per purchase) or 5% in crypto. Regional endpoints cost 10% more than the anthropic endpoint.
- The Anthropic API uses the standard global-endpoint price; US-only inference (inference_geo: "us") is 1.1×. Batch discounts, Priority Tier and long-context premiums (where they apply) are not included.
- Wokey prices move with model supply and demand, and each request is billed at the price when it is processed. Every call’s meters and cost are listed in the console request log.
To use your real usage, the Anthropic Console Usage page or the Wokey console request log both list uncached input, cache reads and writes, and output separately.
Price sources: Anthropic API pricing · Claude plan pricing · OpenRouter Claude endpoint prices · OpenRouter credit fees (FAQ) · Wokey public pricing API
FAQ
How much does Claude Code cost per month on the API?
With this page’s “Claude Code, daily use” preset (100M input tokens a month, 90% cache reads and 5% cache writes, plus 2M output tokens), Claude Sonnet 5 costs about $60.50 a month on the Anthropic API, $63.83 on OpenRouter and $10.89 on Wokey, 82% less than official. Real usage varies widely, so enter your own numbers.
Why does the cache hit rate matter so much on a Claude API bill?
A Claude cache read costs a small fraction of uncached input (10% for Claude Sonnet 5), while a 5-minute cache write costs 1.25× input and a 1-hour write 2×. Claude Code resends a long context every turn, so cache hits cut input cost sharply; when the cache keeps expiring, write charges push the bill up instead.
Is Claude on OpenRouter more expensive than the Anthropic API?
OpenRouter’s anthropic endpoint charges the same token price as Anthropic, with no markup. The difference is the top-up fee: 5.5% when buying credits by card (minimum $0.80), or 5% in crypto. Regional endpoints such as US or EU add 10% to the token price.
Is a Claude Pro or Max subscription cheaper than the API?
Claude Pro is $20 a month and Max starts at $100, with usage metered in rolling 5-hour windows plus weekly caps rather than published token amounts. If your API estimate is under $20, paying per token costs less. If a plan’s allowance covers you and your estimate is well above the plan price, the plan costs less. If you keep hitting limits, run parallel sessions or call Claude from a server, pay per token.
Why is Claude cheaper on Wokey, and is it the official model?
Claude requests go to Anthropic’s official API and the model is never swapped. The saving is on the supply side: supply comes from authorized official subscription quota, which costs less per token than the API list price. With Official verification on, raw passthrough responses carry a Nitro Enclave signed proof that the open-source verifier checks locally.
Will the estimate match my actual bill?
Not exactly. The calculator uses current prices and the usage you enter; Wokey prices move with supply and demand, and each request is billed at the price when it is processed. Tax, batch discounts and long-context premiums are not included. Every call’s meters and cost can be checked in the Wokey console request log.