On flat pricing, every customer carries a hidden threshold: the usage level at which the LLM cost they generate equals the subscription they pay. Below it, they're profitable. Above it, they cost you money on every request. That threshold is break-even usage — and knowing it per plan is the difference between flat pricing that's safe and flat pricing that quietly bleeds you.
What break-even usage is
Break-even usage is the point where a customer's variable cost equals their revenue. On a flat plan the math is simple: break-even = plan price ÷ cost per unit of usage. Pick the unit that drives your bill — output tokens, requests, or total tokens — and the break-even is how many of them a customer can consume before their cost reaches what they pay. Below that line is margin; above it is loss, one token at a time.
On a $29/mo plan served by GPT-4o ($10 / Mtok output), the output-only break-even is about 2.9M output tokens a month. That's the generous ceiling — it ignores input tokens, which push the real break-even lower. A power user running an agent can pass it in days, not weeks.
Why it's the number that predicts losses
Your median customer sits comfortably below break-even, which is why the blended margin looks fine. But usage on a flat plan is wildly uneven, and your heaviest accounts cross the line — with nothing in the pricing to stop them. Break-even usage is what tells you which specific customers are about to flip from profit to loss, before the model invoice lands. It turns 'somewhere we're losing money' into 'these three accounts, this week'.
How to find yours
Three inputs: the plan price, your cost per unit (the model's token rate × the tokens a customer actually uses), and the customer's real usage. Divide price by unit cost to get the break-even level for the plan, then compare each customer's usage against it. The fastest way is to plug a plan price, a model and a usage level into the free calculator and read the point where margin hits zero — no spreadsheet required.
See the rate behind each model
What to do once you know it
- Set a usage cap at or below the plan's break-even — a hard monthly limit, not a vague 'fair use'.
- Nudge customers approaching break-even to a higher tier before they cross it, not after.
- Route heavy accounts to a cheaper model to push their break-even further out.
- Add metered overage beyond an included allowance so the tail pays for what it uses.
- Alert the moment an account crosses break-even, so you act in days instead of at month-end.
The free calculator shows exactly where a customer turns unprofitable for any price, model and usage — then connect Stripe and your LLM cost with MarginWard to track every customer against their break-even automatically. Figures use indicative public rates and illustrative usage.
Go deeper