Free stress test

Put your AI startup on the table.

Enter one period of revenue and usage assumptions. The autopsy shows what remains after refunds, payment fees, and priced model calls—then points to the clearest pressure without pretending it is an audit.

The assumptions

Enter every money amount in USD and every input for the same period. Model rates are in USD; no currency conversion is performed. Nothing here is connected to your real product yet.

Currency (USD only)

Gross revenue

Refunds

Payment processor fees

Active users

Plan price per user

Model

Model calls per user

Extra retry calls (%)

Uncached input tokens / call

Cache-read tokens / call

Output tokens / call

5-minute cache writes / call

1-hour cache writes / call

Replace assumptions

Use the same formula on your real product.

Connect analytics, billing, and server-side model usage. Missing data remains missing; observed results never inherit these estimates.

Use the SDK acceptance checklist to verify the first real event, or set a measurable AI trial budget before offering free usage.

Replace assumptions with real data

Calculation method

Estimated AI cost = users × calls per user × (1 + extra retry percentage / 100) × token cost per attempt. Token cost is the sum of uncached input, cache reads, supported cache writes, and output tokens, each multiplied by its USD rate per million tokens and divided by 1,000,000.

Observed-input contribution = gross revenue − refunds − payment fees − estimated AI cost. Contribution margin divides that result by revenue after refunds and fees, not by gross sales. The plan-price input is used for the pressure comparison; it does not replace entered revenue.

Worked example · synthetic, USD

100 active users × 200 calls × 1.10 retry multiplier = 22,000 attempts. With GPT-4.1 mini, 2,000 uncached input tokens at $0.40 and 1,000 output tokens at $1.60 per million cost $0.0024 per attempt. No cache tokens are included. Estimated AI cost = $52.80. Revenue $1,280 − refunds $40 − fees $38 = $1,202; contribution = $1,149.20.

This is a stress test, not accounting profit. It excludes any hosting, payroll, taxes, discounts or other costs you did not enter. Unknown fees and usage must not be entered as zero. All monetary inputs must be USD; old non-USD scenario links are rejected without an exchange-rate assumption.

Price basis and limits

These are dated catalog rates, not a live price feed. OpenAI and Anthropic entries were recorded on 2026-07-28; BytePlus Seed 2.0 Lite on 2026-07-13. Check the selected model's source and verification date shown with the result before budgeting. Re-running this page does not update those dates.

Only supported token buckets are priced. Models without a separate cache-write rate reject cache-write input. Seed 2.0 Lite rejects more than 128,000 input-side tokens per request because the higher pricing tier is not modeled. No FX conversion is applied.

Provider references: OpenAI pricing, Anthropic pricing, and how Grow or Die versions prices. This tool is an estimate, not a reconciled provider bill.