Tools · Calculator
Claude Token Cost Calculator
Turn tokens and messages into a monthly and annual spend estimate. Paste or type your volume, pick a model, factor in cache reads and batch, then see a number you can act on.
Not affiliated with Anthropic. Claude Token Optimization is an independent site and audit service. Claude and related product names are trademarks of Anthropic PBC. Plan prices, model rates, and usage limits change; always confirm on Anthropic’s own docs.Figures last verified 17 September 2026· Anthropic pricing and usage references
Estimator mode
Your Claude usage
What this calculator does
This calculator is an API spend estimator. Give it a model, monthly input and output volume, cache-read share and Batch API share. It multiplies your volume by per-million-token rates and returns monthly and annual totals.
Subscription cards are included only as rough context. They are not the same billing system as API token charges. For planning context, read API pricing, usage limits and the optimization playbook. For limit monitoring habits, use the Claude usage tracker.
Default rate table
API rates are shown in USD per million tokens. Cache-write premiums are intentionally left out of this v1 calculator. The cache field is a simplified share of input billed at the read rate.
| Model | Input / MTok | Output / MTok | Cache read / MTok |
|---|---|---|---|
| Haiku 4.5 | $1 | $5 | $0.10 |
| Sonnet 5 | $2 | $10 | $0.20 |
| Opus 5 | $5 | $25 | $0.50 |
| Fable 5.1 | $10 | $50 | $0.25 |
| Plan | Rough card | Notes |
|---|---|---|
| Pro | $20 / month | Roughly $20/month, with an annual-equivalent plan commonly shown near $17/month. |
| Max 5x | $100 / month | From $100/month for higher session capacity than Pro. |
| Max 20x | $200 / month | A higher Max tier for heavier usage context. |
| Team standard seat | $25 / month | Shown as $25/seat/month for a clear monthly card estimate. |
| Enterprise seat | $20 / month | Rough seat context only. Usage can be separate. |
How the estimate works
- Convert monthly input and output into millions of tokens.
- Split input into standard input and cache-read input using your cache-read share.
- Apply Batch API share to uncached input and all output at half the standard rate.
- Add the line items into a monthly estimate.
- Multiply by 12 for the annual view.
How to use the results
A high annual number usually means volume plus structure. Ask whether you are re-pasting the same brief every turn, letting tool loops get chatty, or using a heavier model than the task needs.
Optimize first when the estimate is high and you recognize waste patterns: re-pastes, long threads, model overkill or oversized outputs. The audit is built for that.
Buy more capacity when you are already lean, still hitting limits or needing more throughput, and the work is valuable. Use usage limits and API pricing as planning frames.
FAQ
How accurate is this?
It is an estimator, not a bill. Accuracy depends on your token counts and whether you model cache and batch. Always re-check Anthropic’s pricing page before budgeting.
Does caching always lower cost?
Cache reads are cheaper than base input. Cache writes cost more than base input. Caching pays off when you reuse the same prefix enough times within the time-to-live window.
Can I use this for Claude.ai Pro or Max?
Only as a rough sense of what this volume would cost on the API. Pro and Max include usage under plan limits, including a rolling five-hour session window and weekly limits on paid plans. Chat, Claude Code and Cowork draw from the same usage pool on those plans.
Where do I see official usage?
In Claude settings, use Settings > Usage for session and weekly progress on supported plans. For a manual habit, use the usage tracker.
Who runs the audit CTA?
Justin Davis operates an async paste-and-pay Claude token audit. You paste recent prompts and replies; you get a written map of waste and a change list. No call required. Details live on the audit page, with privacy notes and a sample report.