Model LLM costs · check margins · avoid selling below cost at scale
| Tier | Price/mo | AI Cost/user | Gross Margin | LTV | Status |
|---|
Select your primary model to auto-fill costs.
| Cost Component | $/user/mo | % |
|---|
Common questions about AI SaaS pricing, LLM economics, and how this calculator works.
Traditional SaaS targets 70–80% gross margin. AI SaaS typically runs lower — 55–70% is considered healthy — because LLM token costs are a direct cost of goods sold (COGS) that scales with every user action.
Below 40%, you are likely at an unsustainable unit economics floor and will struggle to raise institutional funding or reach profitability at scale.
Formula:
(input_tokens × input_rate + output_tokens × output_rate) × requests × (1 − cache_rate)
Example with GPT-4o Mini at $0.15/1M input, 800 input + 400 output tokens, 120 requests/month, 30% cache: roughly $0.006/user/month. Use the AI Cost Model tab for live calculations against any model.
The benchmark is 3:1 or higher — every dollar spent acquiring a customer should return at least $3 in lifetime value. Below 1.5:1 signals the business burns acquisition spend faster than it recovers it.
Early-stage companies often run below 3:1 while finding product-market fit, but investors expect improvement before Series A.
Providers like Anthropic and OpenAI cache reusable prompt prefixes (system instructions, documents) so repeated calls skip re-tokenizing them — cutting input token costs by 50–90% on the cached portion.
Even a 30% cache hit rate meaningfully reduces your monthly LLM bill. At scale with heavy system prompts, it can be your single biggest cost lever.
Per-seat gives predictable MRR and is easiest to sell, but can feel unfair to light users and under-captures value from power users.
Usage-based aligns cost with value and expands naturally, but creates lumpy revenue. Most AI SaaS eventually lands on hybrid: a base seat fee with a usage allowance, plus overage — protecting your margin floor while letting big users pay more.
For B2B SaaS, 1–2% monthly churn (12–22% annually) is acceptable early-stage. Best-in-class companies target under 0.5%/month.
The LTV math is brutal at high churn: at 5%/month a customer stays an average of 20 months; at 1% they stay 100 months — a 5× LTV difference from the same ARPU.
No. Every calculation runs entirely in your browser. Nothing is
transmitted to any server, no account is required, and nothing persists beyond your own
browser's localStorage — used only to remember your last input values
across refreshes.
Open DevTools → Network tab and you'll see zero outbound requests when interacting with the calculator.
CAC payback is how many months to recover your customer acquisition cost, calculated as
CAC ÷ ARPU.
Under 12 months is excellent. 12–24 months is acceptable for B2B. Over 24 months is a warning sign — you're funding a long wait before each customer becomes profitable, which strains cash as you grow.