OpenAI
Reasoning model (o-series) cost & gross margin per customer
Reasoning models (o-series style) trade cost for deeper thinking, with the highest token prices here. A few heavy reasoning customers can quietly turn unprofitable on a flat subscription — exactly the case to watch.
Reasoning models carry the highest token prices here, and they also tend to generate long internal chains of thought — a double hit. A handful of heavy reasoning users can quietly erase the margin from dozens of light ones.
Input
$5 /Mtok
Output
$20 /Mtok
Margin per customer by usage & plan price
How Reasoning model (o-series) margin holds up as a customer's usage rises, across common subscription prices.
| Usage / mo | LLM cost | $19/mo | $29/mo | $49/mo | $79/mo |
|---|---|---|---|---|---|
| Light | $1.10 | 94.2% | 96.2% | 97.8% | 98.6% |
| Typical | $5.50 | 71.1% | 81% | 88.8% | 93% |
| Heavy | $22.00 | -15.8% | 24.1% | 55.1% | 72.2% |
| Power user | $90.00 | -373.7% | -210.3% | -83.7% | -13.9% |
Margin % per customer at each plan price. Token prices indicative, as of 2026-06.
Of the $5.50 a typical customer costs on Reasoning model (o-series), output tokens are $3.00 (55%) and input $2.50. Output is priced at $20/Mtok — 4× the input rate — so the more your product generates per request, the faster a customer's margin slips.
Worked example
Take a power user on your $49/mo plan sending 8M input / 2.5M output tokens a month. On Reasoning model (o-series) that's $90.00 in tokens — you're losing $41.00 on that single customer every month (-83.7% margin).
How to keep Reasoning model (o-series) profitable
- Trim and cache input context — long system prompts and re-sent chat history are pure, repeated cost.
- Cap output length and stop generation early where you can: at roughly 4× the input price, every extra generated token is where Reasoning model (o-series) hurts most.
- Route easy requests to a cheaper model and reserve Reasoning model (o-series) for the hard ones that actually need it.
- Set a per-customer margin alert so one heavy user can't quietly slip into the red unnoticed.
When to choose Reasoning model (o-series)
Choose a reasoning model only for tasks that genuinely need it, and put it behind your higher-priced tiers or usage caps — its cost can dwarf the rest of your stack on a flat plan.
FAQ
- How much does Reasoning model (o-series) cost per customer?
- At a typical 500k input / 150k output tokens per customer per month, Reasoning model (o-series) costs about $5.50 per customer (input 5/Mtok, output 20/Mtok).
- Is Reasoning model (o-series) profitable for a $49/mo AI SaaS?
- At typical usage, yes — margin is about 88.8% ($43.50 per customer). It erodes as usage rises; heavy and power users are where Reasoning model (o-series) can turn unprofitable.
- What's a good gross margin for an AI SaaS using Reasoning model (o-series)?
- Most AI products target a 60–80% gross margin. With Reasoning model (o-series) at typical usage you're around 88.8% on a $49 plan — comfortable — but your blended margin depends on the heavy users, which is the number worth watching.
- At what usage does Reasoning model (o-series) stop being profitable on a $29 plan?
- Around 2.6M input / 0.8M output tokens a month. Past that point, a $29 customer costs you more than they pay.
- How do I reduce Reasoning model (o-series) cost per customer?
- Cut output tokens first (they're the priciest), cache or trim input context, route easy requests to a cheaper model, and watch the break-even point — around 4.5M input / 1.3M output tokens a $49 customer stops being profitable.
Compare this model
Other models
Key terms