Skip to content
FinCalc Hub

AI API Cost & SaaS Unit Economics Calculator (2026)

Estimate what an AI feature will cost to run each month, compare token pricing across Claude, GPT, Gemini and DeepSeek models side by side, and see your profit margin and break-even user count at your subscription price.

Data as of November 2026

Measure prompts in
$/mo
$/mo

Monthly profit on Claude Sonnet 4.5

$18,263

91.3% margin on $20,000 revenue

AI API cost / month

$1,238

150,000 requests

AI cost per user

$1.24

$0.0083 per request

Break-even users

27

to cover $500 fixed costs

Contribution / user

$18.76

price minus AI cost
Where subscription revenue goes
Profit: $18,263AI API cost: $1,238Fixed costs: $500
  • Profit$18,26391%
  • AI API cost$1,2386%
  • Fixed costs$5003%

Monthly cost by model

Model$ / 1M in · outMonthlyMargin
$0.28 · $0.42$5797%
$0.25 · $2.00$14897%
$0.30 · $2.50$18497%
$1.00 · $5.00$41395%
$1.25 · $10.00$74194%
$1.25 · $10.00$74194%
$3.00 · $15.00$1,23891%
$5.00 · $25.00$2,06387%

List prices, no caching or batch discounts. 30-day month.

How AI API pricing works

Large language model APIs bill per token, with separate prices for the tokens you send (input) and the tokens the model generates (output). Your monthly bill is driven by four numbers: active users, requests per user, average prompt size and average response size. This calculator multiplies them across a 30-day month for each model.

Model cost comparison: 10,000 users

Reference scenario: 10K monthly active users making 5 requests a day with 750-token prompts and 400-token responses (1.5M requests per month). Costs range from $567 on DeepSeek V3.2 to $20,625 on Claude Opus 4.5.

ModelProviderInput / 1MOutput / 1MContextMonthly costPer user
DeepSeek V3.2DeepSeek$0.28$0.42128K$567$0.06
GPT-5 miniOpenAI$0.25$2.00400K$1,481$0.15
Gemini 2.5 FlashGoogle$0.30$2.501M$1,838$0.18
Claude Haiku 4.5Anthropic$1.00$5.00200K$4,125$0.41
GPT-5OpenAI$1.25$10.00400K$7,406$0.74
Gemini 2.5 ProGoogle$1.25$10.001M$7,406$0.74
Claude Sonnet 4.5Anthropic$3.00$15.00200K$12,375$1.24
Claude Opus 4.5Anthropic$5.00$25.00200K$20,625$2.06

Standard list prices per million tokens, before prompt caching, batch discounts or long-context surcharges. Prices change often; check each provider’s pricing page before you commit.

Unit economics: margin and break-even

  • Revenue = active users × subscription price.
  • Profit = revenue − AI API cost − fixed monthly costs (hosting, tools, salaries you allocate).
  • Contribution per user = price − AI cost per user.
  • Break-even users = fixed costs ÷ contribution per user. If AI cost per user exceeds your price, you never break even.

Frequently asked questions

How do I calculate the cost of an LLM API?

Multiply monthly requests by average input tokens and by the model’s input price per million tokens, do the same for output tokens with the output price, and add the two. Monthly requests are monthly active users × requests per user per day × 30.

How many tokens is a word?

In English, one word averages about 1.33 tokens, or roughly 750 words per 1,000 tokens. Code, non-English text and unusual formatting use more tokens per word, so measure real prompts with the provider’s tokenizer before launch.

Why are output tokens more expensive than input tokens?

Generating each output token requires a full forward pass of the model, while input tokens are processed in parallel. Providers therefore price output tokens at 2 to 8 times the input rate, so long responses usually dominate the bill.

How can I reduce AI API costs in my SaaS?

Route simple requests to a smaller model, cache repeated system prompts (prompt caching can cut input costs sharply), cap maximum output length, use batch APIs for non-urgent work, and set per-user usage limits on lower pricing tiers.

What is a healthy gross margin for an AI SaaS product?

Traditional SaaS targets 75% to 85% gross margin. AI-native products often run at 50% to 70% because inference is a real variable cost. If AI cost per user is more than about a third of your price, revisit model choice or pricing.