AI token cost calculator
The cost of an AI API comes down to four numbers: requests per month, input tokens, output tokens and the model's price. The starting values are the example from our article on how to calculate the cost of an AI project: a support chatbot with 300,000 requests per month. You can also call it a simulator: the numbers change as you move the sliders.
A thousand users making ten requests a day is 300,000 a month.
Question, instructions and context. 1 token ≈ 0.75 words.
The generated answer.
For peaks and longer answers: 20% to 30%.
| Model | Input / output per million | Cost per month |
|---|---|---|
| Claude Fable 5.1 | $10.00 / $50.00 | $6,000 |
| Claude Opus 5 | $5.00 / $25.00 | $3,000 |
| Claude Sonnet 5 | $2.00 / $10.00 | $1,200 |
| Claude Haiku 4.5 | $1.00 / $5.00 | $600 |
| GPT-5.5 | $5.00 / $30.00 | $3,450 |
| GPT-5.6 sol | $4.00 / $20.00 | $2,400 |
| GPT-5.4 | $2.50 / $15.00 | $1,725 |
| GPT-5.6 terra | $2.00 / $12.00 | $1,380 |
| GPT-5.4 nano | $0.20 / $1.25 | $143 |
| GPT-5.6 luna | $0.20 / $1.20 | $138 |
| Gemini 3.1 Pro | $2.00 / $12.00 | $1,380 |
| Gemini 3.5 Flash | $1.50 / $9.00 | $1,035 |
| Gemini 3.8 Flash | $0.75 / $3.75 | $450 |
| Gemini 2.5 Flash-Lite | $0.10 / $0.40 | $51.00 |
List prices in US dollars, without caching (up to 90%) or batch (50%) discounts, checked on the providers' official pages on 16 September 2026 (links under Sources). They change often: always check before deciding.
Indicative estimate: real values depend on your process and your data.
How it is calculated
Monthly cost = requests × (input tokens × input price + output tokens × output price) ÷ 1,000,000.
Prices are per million tokens and output is almost always about five times the input price. That is why the answer weighs more on the bill than the question, even with fewer tokens.
The safety margin covers usage swings, long conversations and answers longer than expected. In a proposal, always give a range and the volume it is based on.
Sources
Frequently asked questions
What is a token?
It is the unit models use to read and write text. On average, a token is about 0.75 words, and a million tokens is about 750,000 words.
How do I cut the cost without losing quality?
The biggest lever is choosing the right model for each task: simple tasks rarely need the top tier. Then prompt caching for repeated context, batch for work that doesn't need an instant answer, and limits on answer length.
Why does the same volume cost so differently from model to model?
Because the price per token varies by more than a hundred times between the cheapest and the most expensive model. Within one provider, the gap between models matters more than the gap between providers.
Prices are in dollars. What is that in euros?
Providers bill in US dollars. Convert at the day's exchange rate; since prices and rates change, treat the result as an estimate, not a quote.
Is this a calculator or a simulator?
Here they are the same thing: change the values and see the result instantly, with nothing to install and no details to leave. If you were looking for an AI cost simulator, this is it.
Other calculators
Want help putting this into practice?
In a free 30-minute assessment we look at your operation and show you where the biggest return is.
Get a free assessment