# Eval run cost estimator

Estimate the API bill for one evaluation run from its task count, token use, model prices, and likely retries.

Use the tool on the [web page](https://aibenchmarkindex.com/tools/eval-run-cost-estimator). It works without JavaScript.

## How to use it

1. Enter number of tasks.
2. Enter average input tokens per task.
3. Enter average output tokens per task.
4. Enter input price per million tokens.
5. Enter output price per million tokens.
6. Enter retry allowance.
7. Read the result.

## How it works

We multiply tasks by average input and output tokens, price each total at the per-million-token rate, and add the two charges.

The low estimate assumes one call per task. The high estimate applies the retry allowance to the entire base cost. It does not include storage, judge-model calls, tool calls, taxes, or provider discounts.

> This is an estimate. Provider rounding, caching, batch discounts, taxes, and failed requests can change the bill.

## Questions

### Does this include an LLM judge?

No. Add the judge's tokens as another run or include them in the per-task averages.

### Why is this a range?

The low end assumes no retries. The high end adds the retry percentage you enter.

> **Check before you compare.** Benchmark methods and leaderboards change. Follow the cited primary source before using a result in a decision.