AI SaaS margin calculator
Enter your seat price, how much each user uses the product, and the model behind it. You get cost per user, gross margin per plan, and the number of requests where a user starts to lose you money.
Result
One request costs $0.0112 on Claude Sonnet 5.5. Infra is spread over 500 paying users.
| Plan | AI cost / user | Stripe / user | Infra / user | Cost / user | Gross margin | Break-even requests | Requests at 60% margin |
|---|---|---|---|---|---|---|---|
| Starter $20.00 | $3.35 | $1.02 | $0.80 | $5.17 | 74.2% | 1,630 | 554 |
| Pro $60.00 | $16.73 | $2.46 | $0.80 | $19.99 | 66.7% | 5,088 | 1,860 |
Blended gross margin across all plans: 71.0%
Gross margin here is price minus model tokens, Stripe fees, and infra. It leaves out support, refunds, taxes, and free users. The uncached share of input is billed at the base input price. Cache writes can cost more (1.25x on Claude), so a low hit rate costs a little more than shown. Token counts for the same text differ between providers.
The default scenario, worked out
Claude Sonnet 5.5, 3,000 input and 800 output tokens per request, 50% of input served from the prompt cache, $400a month of infra, and Stripe at 3.6% + $0.30 per charge. One request costs $0.0112.
| Plan | Users | Requests / user | Cost / user | Gross margin | Break-even requests |
|---|---|---|---|---|---|
| Starter at $20.00 | 400 | 300 | $5.17 | 74.2% | 1,630 |
| Pro at $60.00 | 100 | 1,500 | $19.99 | 66.7% | 5,088 |
Blended gross margin: 71.0%.
The Starter plan on every model
Same tokens, same $20.00 price, 300 requests per user. Only the model changes.
| Model | Input / output per 1M | AI cost / user | Gross margin | Break-even requests |
|---|---|---|---|---|
| Claude Opus 5.5 | $4 / $20 | $6.69 | 57.4% | 815 |
| Claude Fable 5.1 | $10 / $50 | $16.61 | 7.8% | 328 |
| Claude Sonnet 5.5 | $2 / $10 | $3.35 | 74.2% | 1,630 |
| Claude Haiku 5.5 | $0.1 / $0.5 | $0.17 | 90.1% | 32,176 |
| Claude Sonnet 4.6 | $3 / $15 | $5.09 | 65.5% | 1,072 |
| GPT-6 Astra | $10 / $50 | $16.95 | 6.2% | 321 |
| GPT-6.1 Sol | $2 / $10 | $3.35 | 74.2% | 1,630 |
| GPT-6 Luna | $0.1 / $0.5 | $0.17 | 90.1% | 32,176 |
| GPT-5.6 Sol (promo price) | $4 / $20 | $6.78 | 57.0% | 804 |
| GPT-5.6 Terra | $2 / $12 | $3.87 | 71.5% | 1,409 |
| GPT-5.6 Luna | $0.2 / $1.2 | $0.39 | 89.0% | 14,093 |
| GPT-5.4 mini | $0.75 / $4.5 | $1.45 | 83.6% | 3,758 |
| Gemini 4 Argon (2026 price) | $2 / $10 | $3.35 | 74.2% | 1,630 |
| Gemini 4 Argon (from 2026-12-01) | $4 / $20 | $6.69 | 57.4% | 815 |
| Gemini 3.1 Pro (preview) | $2 / $12 | $3.87 | 71.5% | 1,409 |
| Gemini 3.8 Flash (2026 price) | $0.75 / $3.75 | $1.27 | 84.5% | 4,290 |
| Gemini 3.8 Flash (from 2027-01-01) | $1.5 / $7.5 | $2.54 | 78.2% | 2,145 |
| Gemini 3.5 Flash-Lite | $0.3 / $2.5 | $0.75 | 87.2% | 7,286 |
| Gemini 3.1 Flash-Lite | $0.25 / $1.5 | $0.48 | 88.5% | 11,274 |
Prices from the official pages of Anthropic,OpenAI, andGoogle, and Stripe fees fromstripe.com/pricing, all checked 2026-10-13. Different tokenizers turn the same text into different token counts, so treat cross-provider rows as approximate.
How it works
Cost per request is input tokens at the input price, with the cached share at the cache price, plus output tokens at the output price. Cost per user adds requests per month times that, one Stripe charge, and an even share of your infra bill. Gross margin is price minus that cost, divided by price.
Break-even requests is the monthly usage where a user's token bill eats everything left after Stripe and infra. Above that line, the user costs you money.
What it leaves out
Support time, refunds, free users, sales tax, and model retries are not in the formula. Neither are long-context surcharges or Batch API discounts. If a plan only clears your target with nothing to spare, those will push it under.
The worked examples and the reasoning behind each input are in our guide,how to price an AI SaaS.
Questions clients ask
Short answers, no sales language. If yours isn't here, ask an engineer directly.
Ask us anythingWhat gross margin should an AI SaaS aim for?
There is no single right number. Bessemer's State of AI 2025 report found its fast, durable AI startups averaged about 60% gross margin, and the very fastest averaged about 25%. Classic SaaS often runs at 80% or more. Pick a line you can live with and set it as the warning threshold here.
Why does the calculator include Stripe fees?
Because they are a real cost of every charge. Stripe's standard US price is 2.9% + $0.30 per card charge, and Stripe Billing adds 0.7% of billing volume. On a $20 plan that is $1.02 a month, about 5% of the price, before you spend a cent on tokens.
How do I find my tokens per request?
Read the usage object your model API returns on each response. It reports input, cached input, and output tokens. Log it for a week of real traffic and use the average, then check the 90th percentile user too.
Where do the model prices come from?
From each provider's official pricing page, checked 2026-10-13: Anthropic, OpenAI, and Google. Prices are standard tier, short context, global routing. Each row in the tool links to its source.
Building an AI product?
We build AI SaaS products end to end, billing and usage limits included. An engineer replies, not a sales team.