LLM Pricing Calculator
Vendor pricing pages quote dollars per million tokens. Nobody thinks in millions of tokens — they think in requests. Put your workload in and see the monthly bill across 20 priced models, with cached input and batch discounts applied where the vendor offers them.
| Model | Per month | Per 1k requests | Per year |
|---|
Unverified figures · Model index · Compare two models
How the estimate is worked out
Every figure is list price, in US dollars, per million tokens, taken from the model index. The arithmetic is deliberately simple, because the complicated part of an AI bill is never the multiplication:
- Input — requests × input tokens, split between the standard rate and the cached rate by the share you set. A vendor with no published cached rate is charged at its standard rate.
- Output — requests × output tokens at the output rate. Output is where the money goes: most vendors charge three to five times more for it.
- Batch — the share you mark as asynchronous is charged at the vendor's batch rate where one is published, and at the standard rate where it is not.
What it does not include: fine-tuning, storage, image or audio tokens, rate-limit headroom, the reasoning tokens some models bill as output, or the requests you retry. Treat the result as the floor of your bill rather than the number.
Prices move. Check the vendor's own page before you commit a budget — each family in the model index links to its source.