Model your AI SaaS margins before your OpenAI bill eats them.

Name: LLM Margin
Author: LLM Margin

LLM Margin is a margin simulator for founders shipping AI products. Enter your pricing, token usage, and MAU — get gross margin, cost-per-user, and breakeven in seconds.

·See your true gross margin % after LLM costs — not the back-of-napkin version your board deck uses.
·Spot power-user risk before one heavy customer wipes out margin on the other 500.
·Find the MAU number where you actually break even at your current pricing.

Pressure-test your AI unit economics before raising or repricing.

Your SaaS Pricing

Monthly price per user (ARPU) ($)

What you charge per user per month. If you have tiers, use your most common paid tier.

Monthly Active Users (MAU): 1,000

Fixed monthly costs ($)

Hosting, tools, your own salary if applicable. Used for breakeven calculation.

Your AI Usage

LLM Model

AI calls per user per day: 5

How many times does your app call the LLM per active user per day?

Avg input tokens/call

System prompt + user message + history

Avg output tokens/call

Typical AI response length

≈ $0.011500 per call · $1.72 per user/month · $1,725/mo at 1,000 users

Gross Margin

Healthy

83.0%

Software target: 70–80%. AI SaaS avg: 45–60%.

MRR

$19,000

1,000 users × $19.00/mo

AI COGS / month

$1,725

$1.7250/user/mo · OpenAI: GPT-5.5

Breakeven MAU

127

Min paying users needed to cover fixed costs.

Gross Margin vs. Scale

Healthy margins. Your AI costs are under control at current scale.

Prices sourced from OpenRouter and updated automatically. Always verify with your provider before making pricing decisions. Estimates do not include volume discounts or enterprise rate cards.

LLM Margin is a free LLM margin calculator built for SaaS founders, indie hackers, and AI startup CTOs who need real answers about AI cost per user, gross margin, and breakeven MAU. Instead of spreadsheets and Reddit threads, plug your OpenAI, Anthropic, or open-source LLM pricing into the SaaS margin simulator and get cost-per-user, power-user risk, and monthly burn in one view. If you're pricing an AI product or defending margin to investors, this is the fastest way to stress-test the model.