Inference cost
What will the model calls cost once in production?
Requests, tokens, model tier, cache and growth: the monthly bill, its trajectory, and what a lower tier would save.
- ≈ 3 min
- Free calculation · AI commentary on request
- You leave with: The monthly bill and its two-year trajectory
The workload
For a per-user cost. Optional.
System prompt + context + question. A detailed system prompt is 500–2,000 tokens.
A paragraph ≈ 100 tokens. A page ≈ 500.
Share of input context served from the prompt cache. A fixed system prompt plus shared context easily reaches 60%.
The bill
The figures above are computed and free. If you also want them read back to you — what they mean, what is specific to your sector, where to start — that is produced on request.
Month by month
With the growth entered. The line that climbs fast is the one discovered in the third quarter.
The detail
| Line | Estimate |
|---|---|
| Recurring | |
Input tokens (uncached) 91M tokens/month × €2.5/M. | €137 €44 — €219 |
Cached input tokens 40% of context served from cache, at −75%. | €23 €7 — €36 |
Output tokensoften omitted 24M tokens/month × €10/M. Several times the input price: answer length is a budget line. | €243 €97 — €389 |
Sensitivity
The bill depends on several assumptions at once. This shows which one to check first if the result feels uncertain.
Each assumption is moved ±20% on its own, holding the others fixed. The longest bar is the one worth checking first — not necessarily the one that felt most uncertain.
The assumptions
Every figure above rests on these. The ones marked “assumption” have no external source: they are modelling choices, labelled so they can be argued with rather than mistaken for measurements.
- Tier standardPublished
- Mid-tier general models.
- Output price 10 €/MPublished
- Mid-tier output pricing.
- Cache discount −75 %Published
- Providers price cached prompt prefixes at a large discount. A stable system prompt plus a long shared context is the common case where this matters.
- Input share of the bill 34 %Observed
- The rest is output. If this share is small, shortening answers pays more than compressing context.
- Price volatility quarterlyPublished
- Published prices fall regularly and model names change. The tiers persist; their occupants rotate. Check the vendor's grid before contracting.
What the calculation does, what the AI does
Every figure on this page comes from a deterministic algorithm running in your browser: the same input always gives the same result, and nothing is sent to produce it. AI runs only afterwards, to write the commentary from those figures. It produces none of them — the format it fills has no numeric fields — and any sentence quoting a value the calculation did not produce is removed before display.
Where to go next
See all 19 tools →- Cost calculator
What does our AI transformation actually cost, and when does it break even?
≈ 3 min · A line-by-line budget, the return and the break-even month - Use-case prioritisation
Of all our ideas, which one first?
≈ 5 min · Your use-case list ranked, on a value/effort plane - Build or buy
Should we buy this tool or build it ourselves?
≈ 3 min · Both options costed and the point where building pays - Right-size licences
How many seats are we paying for nothing?
≈ 2 min · The saving available at renewal, and the tiered arrangement - Internal team or external
From when is hiring cheaper than buying days?
≈ 3 min · The crossover month and how many hires are justified - Automate a process
Is this process worth automating, and how many months until it pays back?
≈ 3 min · A signable automation case, with a proposed decision