aiapicostcalcUnit economicsPricesBrandsCompareBenchmarksSkills & PluginsFind my modelGuidesChanges

AI SaaS unit economics calculator

Short answer: Cost per user per month = calls per user per day × 30 × cost per call. Minimum subscription price = cost per user ÷ (1 − target gross margin). Enter your own token counts and margin below; the result is the lowest price that keeps that margin on the verified list rates.

Cost per user per month = calls per user per day × 30 × cost per call, and cost per call = (input tokens × input rate + output tokens × output rate) ÷ 1,000,000. In the calculator's default case — 2,000 input and 500 output tokens, 1,000 users at 5 calls a day and an 80% margin — the bill comes to 150,000 calls a month and the page shows a $6.75 minimum price per user.

Last verified 2026-11-18 · Sources: official provider pricing pages, copied into src/prices.json (checked 2026-11-18) · Report an error

1
2
Presets

Preset token counts are starting assumptions; replace them with your logged averages.

3
Cost per call$0.0090
Per 1,000 calls$9.00
Calls per month150,000
Model bill per month$1,350
Cost per user per month$1.35
Minimum price per user (80% margin)$6.75

$2 in · $10 out per 1M

All models, this workload

ModelPer callPer 1K callsMin price / user
Ministral 3 3B$0.00025$0.2500$0.1875
Ministral 3 8B$0.000375$0.3750$0.2813
Gemini 2.5 Flash-Lite$0.0004$0.4000$0.3000
GPT-6 Luna$0.00045$0.4500$0.3375
Claude Haiku 5.5$0.00045$0.4500$0.3375
Ministral 3 14B$0.0005$0.5000$0.3750
GPT-4o mini (deprecated)$0.0006$0.6000$0.4500
Mistral Small 4$0.0006$0.6000$0.4500
GPT-5.6 Luna$0.0010$1.00$0.7500
GPT-5.4 nano (deprecated, retires 2027-04-01)$0.00103$1.03$0.7688
Codestral 25.08$0.00105$1.05$0.7875
DeepSeek Flash (V4.1)$0.0012$1.20$0.9000
Gemini 3.1 Flash-Lite$0.00125$1.25$0.9375
GPT-5 mini (deprecated, retires 2026-12-11)$0.0015$1.50$1.13
Gemini 3.5 Flash-Lite$0.00185$1.85$1.39
Gemini 2.5 Flash$0.00185$1.85$1.39
Mistral Large 4$0.00245$2.45$1.84
Grok Build 0.1$0.0030$3.00$2.25
Gemini 3.8 Flash$0.00337$3.38$2.53
Gemini 3.6 Flash$0.00337$3.38$2.53
GPT-5.4 mini$0.00375$3.75$2.81
Grok 4.3$0.00375$3.75$2.81
o4-mini (shut down 2026-10-23)$0.0044$4.40$3.30
GPT-6.1 Terra$0.0045$4.50$3.38
Claude Haiku 4.5$0.0045$4.50$3.38
DeepSeek V4 Pro$0.00462$4.62$3.47
Grok 4.7$0.0070$7.00$5.25
Gemini 2.5 Pro$0.0075$7.50$5.63
GPT-4.1 (deprecated)$0.0080$8.00$6.00
GPT-6.1 Sol$0.0090$9.00$6.75
GPT-6 Sol$0.0090$9.00$6.75
Claude Sonnet 5.5$0.0090$9.00$6.75
Claude Sonnet 5$0.0090$9.00$6.75
Gemini 4 Carbon$0.0090$9.00$6.75
Gemini 4 Argon$0.0090$9.00$6.75
GPT-5.6 Terra$0.0100$10.00$7.50
Gemini 3.1 Pro Preview$0.0100$10.00$7.50
GPT-5.4$0.0125$12.50$9.38
Claude Sonnet 4.6$0.0135$13.50$10.13
GPT-5.6 Sol$0.0180$18.00$13.50
Claude Opus 5.5$0.0180$18.00$13.50
GPT-5.5 Rosalind Research$0.0225$22.50$16.88
GPT-5.5 Rosalind Discovery$0.0225$22.50$16.88
Claude Opus 5$0.0225$22.50$16.88
GPT-5.5$0.0250$25.00$18.75
Claude Fable 5.2$0.0360$36.00$27.00
Claude Mythos 5.2$0.0360$36.00$27.00
GPT-6 Astra$0.0450$45.00$33.75
GPT-6 Rosalind$0.0450$45.00$33.75
Claude Fable 5.1$0.0450$45.00$33.75
Claude Mythos 5.1$0.0450$45.00$33.75
Claude Fable 5$0.0450$45.00$33.75
Claude Mythos 5$0.0450$45.00$33.75
GPT-5.6 Cyber$0.0625$62.50$46.88
GPT-6 Cyber$0.0900$90.00$67.50

o4-mini is shut down; its last published list rate is kept for comparing old invoices, not as a model you can call today — the model page names its replacement.

How to get good inputs

Log 50–100 real requests. Average the input tokens (system prompt + history + user message) and the output tokens. If a fixed prefix repeats, enter its share as cached %. Many SaaS companies target 70–80% gross margin; the minimum price is cost per user ÷ (1 − margin).

Worked examples with today's prices

Three workloads run through the same arithmetic as the calculator, using the verified list price of each model. Cached input is billed at the model's cached rate; no batch discount.

Support bot: Claude Haiku 4.5

Prices: $1.00 input, $0.10 cached input, $5.00 output per 1M tokens (official page, checked 2026-10-27). Workload: 2,000 input tokens per call (50% cached), 400 output tokens, 3,000 monthly active users, 8 calls per user per day, 80% target gross margin.

  1. Cost per call = (1,000 × $1.00 + 1,000 × $0.10 + 400 × $5.00) ÷ 1,000,000 = $0.00310
  2. Calls per month = 3,000 × 8 × 30 = 720,000
  3. Model bill = 720,000 × $0.00310 = $2,232.00 per month
  4. Cost per user = $2,232.00 ÷ 3,000 = $0.744
  5. Minimum price per user at 80% margin = $0.744 ÷ (1 − 0.80) = $3.72

RAG answer with a cached system prompt: GPT-5.4 mini

Prices: $0.75 input, $0.075 cached input, $4.50 output per 1M tokens (official page, checked 2026-10-27). Workload: 6,000 input tokens per call (60% cached), 500 output tokens, 1,500 monthly active users, 5 calls per user per day, 75% target gross margin.

  1. Cost per call = (2,400 × $0.75 + 3,600 × $0.075 + 500 × $4.50) ÷ 1,000,000 = $0.00432
  2. Calls per month = 1,500 × 5 × 30 = 225,000
  3. Model bill = 225,000 × $0.00432 = $972.00 per month
  4. Cost per user = $972.00 ÷ 1,500 = $0.648
  5. Minimum price per user at 75% margin = $0.648 ÷ (1 − 0.75) = $2.59

Long-form drafting: Claude Sonnet 5.5

Prices: $2.00 input, $0.10 cached input, $10.00 output per 1M tokens (official page, checked 2026-10-27). Workload: 1,500 input tokens per call (0% cached), 2,500 output tokens, 500 monthly active users, 4 calls per user per day, 70% target gross margin.

  1. Cost per call = (1,500 × $2.00 + 0 × $0.10 + 2,500 × $10.00) ÷ 1,000,000 = $0.02800
  2. Calls per month = 500 × 4 × 30 = 60,000
  3. Model bill = 60,000 × $0.02800 = $1,680.00 per month
  4. Cost per user = $1,680.00 ÷ 500 = $3.360
  5. Minimum price per user at 70% margin = $3.360 ÷ (1 − 0.70) = $11.20

The third example shows why output-heavy products need a different price floor: output tokens cost four to six times as much as input on most models, so a drafting tool costs far more per call than a chat bot with the same total tokens. Replace the token counts with your own logged averages before you set a price.

Questions

How is the AI API cost calculated?

(uncached input tokens × input price + cached input tokens × cached price + output tokens × output price) ÷ 1,000,000, using each provider's standard per-1M-token price in USD.

Where do the prices come from?

The official provider pricing pages linked on every model page. Each row shows the date that model was checked; the footer shows the most recent check. If a price cannot be verified it is marked unverified and excluded from totals. The verification rules are on the method page.

What should my AI SaaS charge per user?

Minimum price = model cost per user per month ÷ (1 − target gross margin). With $0.50 cost per user and an 80% margin target you need at least $2.50 per user per month.

Are reasoning tokens billed?

Yes, reasoning or thinking tokens are billed as output tokens, so set output tokens above the visible answer length for reasoning models.

Are batch discounts included?

The calculator uses standard pricing. Each model page shows Batch API prices separately: OpenAI, Anthropic and Google list batch at 50% of standard; xAI lists a 20% batch discount on Grok 4.7 and 4.3 and no batch for Grok Build.

Do all models count tokens the same?

No. Tokenizers differ by provider and generation. Anthropic's model docs say its newer tokenizer fits about 555K English words in 1M tokens, versus about 750K words for older Claude models, so the same text can cost more tokens.

How do I calculate the API cost per user for my AI app?

Multiply calls per user per day by 30 by cost per call to get cost per user per month; cost per call is (input tokens × input rate + output tokens × output rate) ÷ 1,000,000. Then divide cost per user by (1 − target gross margin) for the minimum price. In the default example on this page — 2,000 input / 500 output tokens, 1,000 users, 5 calls a day, 80% margin — cost per user is $1.35 and the minimum price is $6.75.

Price data last checked 2026-11-18 against official sources · how we verify

Price-change alerts

One short email when a provider changes a price or a promo ends. No spam, unsubscribe any time.