AI API Cost Calculator
Paste a prompt or enter token estimates, compare current models, and forecast the real cost per request, month, and year.
What are you creating?
How often will you use it?
Choose a model
Where the cost comes from
Advanced usage assumptions
Use custom token prices
Advanced cost breakdown
Compare model costs
| Model | Per request | 1,000 requests | Monthly | Annual | Relative cost |
|---|
How much usage fits your budget?
How AI API cost estimates work
Input versus output
Providers usually charge separately for the tokens you send and the tokens the model generates. Output tokens are often more expensive.
Cached input
Reusable prompts or context may qualify for lower cached-input rates. Cache creation and storage can still have separate charges.
Batch processing
Some providers discount asynchronous batch jobs. The calculator applies only the discount stored for the selected model.
Why the final bill can differ
Tokenizers, reasoning tokens, retries, tools, search, storage, long-context tiers, regional processing, and platform markups can change actual billing.
Common mistakes—and the fix
Only counting the visible prompt
Add system instructions, conversation history, retrieved documents, tool schemas, and any hidden context your application sends.
Forgetting retries and regenerations
Use the retry controls. A 10% retry rate means roughly 1.1 billed calls for each successful request.
Confusing ChatGPT or Claude subscriptions with API billing
This tool estimates API usage. Consumer subscriptions and API charges are normally separate products.
Picking the cheapest model as the automatic winner
Price is only one factor. Quality, latency, context size, tool support, and reliability may make a more expensive model the better value.
