Work out what each request and each user costs you, what to charge to hit your target margin, and which heavy users will quietly lose you money on a flat price.
Prices are USD per 1M tokens. Pick a preset or enter your own.
How much does a typical paying user use your product?
Updates as you type.
Margin at as usage grows.
| Usage | Requests | Cost / user | Margin | Status |
|---|
Estimates only. Excludes taxes, free-tier users, retries, caching discounts and volume pricing. Verify model prices with your provider.
Unlike classic SaaS, every action in an AI product has a marginal cost. The three things that decide your margin are the cost per request, how many requests each user makes, and how your price scales with that usage. This calculator ties them together so you can see, before launch, whether your pricing holds up.
If the stress test shows heavy users going underwater, you have three common fixes: cap usage with included quotas, charge for overage, or move to credits instead of raw tokens. Our guide to usage-based billing for AI products walks through each option, and the glossary explains cost per customer and overage billing.
Ready to try tiers with included usage and overage? Use the Usage Pricing Plan Builder, compare credits, tokens and flat-rate pricing, or estimate your provider bill with the LLM Cost Estimator.
Cost per request = (input tokens ÷ 1,000,000 × input price) + (output tokens ÷ 1,000,000 × output price). Output tokens are usually several times more expensive than input tokens, so a long response costs far more than a long prompt.
Traditional SaaS often runs at 75-80% gross margin. AI products carry real per-request inference costs, so many land lower. A common rule of thumb is to aim for 50-70% on the inference-heavy part of the product and to leave room for model price changes.
A flat monthly price is profitable only for the average user. Because AI cost scales with usage, a small group of heavy users can consume most of your inference budget. The stress test shows the usage multiple at which a user stops being profitable.
The presets are representative list prices in USD per 1M tokens and providers change them often. Always check your provider's current pricing page and edit the price fields, which the calculator uses as entered.
No. Everything is calculated in your browser. The share link stores your inputs in the URL fragment, which is never sent to our servers.