How does the AI API cost calculator work?

It multiplies the selected Model.sale blended rate by the estimated input plus output/reasoning charged-token total. Official comparisons use documented split input and output rates for the same token mix.

Is the initial reservation an extra request fee?

No. The initial reservation temporarily protects the prepaid wallet and is released after terminal usage settles the final charge.

Workload calculator · rates checked 2026-08-31

Estimate your AI API cost.

Enter a representative input/output mix instead of comparing one headline number. The calculator uses the current Model.sale catalog and shows a dated direct-price reference only where the units are comparable.

ModelInput tokensOutput + reasoning tokens

Model.sale

$0.0108$0.0900 / 1M charged tokens

Official split tariff

$0.0440Source and current rates

Before dispatch

$0.0100Temporary initial reservation; it is not the final charge and is released after settlement.

The Model.sale estimate applies the blended rate to input plus output/reasoning tokens. Cache, tools, retries and terminal usage can change the final total. The official column appears only when a comparable split rate is documented.

Use a real sampleExport an hour or day of token usage. Include retries and long reasoning output instead of estimating from prompt characters.

Compare the same unitsA blended rate is not an input-only rate. Keep the input/output mix, currency and cache policy identical.

Verify after the callThe dashboard records terminal usage, the final charge and released reservation for each completed request.

What the estimate excludes

The calculator does not guess tool-call charges, image/audio units, provider-specific cache writes, failed attempts or quality equivalence. If a workload uses those features, calculate them separately from the linked first-party tariff and confirm the actual ledger after a small request.

For the exact settlement rules, read the pricing methodology. For provider and aggregator trade-offs, open the API comparison.