About CostPerPrompt

AI pricing is deceptively hard to reason about. Rates are quoted per million tokens, output costs more than input, caching and batching change the math by 2×, and prices shift every few weeks as new models launch. Most teams discover their real costs from their first invoice — which is the most expensive way to learn.

CostPerPrompt exists to answer one question before that invoice arrives: "what will this actually cost me?" We track live pricing for 232+ models and — more importantly — turn those rates into real-world answers with calculators that model how APIs are actually billed: growing chat history, cache hits, batch discounts, and input/output asymmetry.

Where the data comes from

Pricing is refreshed automatically from public provider listings (primarily the OpenRouter public model index, cross-checked against vendors' official pricing pages). Every page shows its data refresh date — currently 2026-08-02. If you spot a discrepancy, please tell us; corrections ship within hours.

How estimates are calculated

Our calculators use the standard billing formula (tokens ÷ 1,000,000 × rate, separately for input and output) with documented assumptions shown next to each result. Estimates are planning aids, not quotes: your provider's official pricing page and your own measured token counts are the final word.

Independence

CostPerPrompt is independent and not affiliated with any AI provider. The site is supported by advertising and some affiliate links (always at no extra cost to you, and never affecting how we rank or present prices).