Token count · API cost · no upload
Price a prompt before you send it
Paste the text, enter your provider's per-million rates, and see the token count and cost in dollars and rupees. Useful when you are about to feed a 40-page PDF to a model and want to know what that habit costs at scale.
Count it, cost it
The estimate blends a characters-per-token and a words-per-token rule. The band around the number is the honest part — real tokenizers disagree with each other too.
These two numbers come from your provider's price page. We do not store a price list, because it would be wrong within a month — and a stale price is worse than no price.
0
likely between 0 and 0 tokens
- As prompt input
- $0.00
- ≈ ₹0.00
- As model output
- $0.00
- ≈ ₹0.00
0 characters · 0 words · 0 lines
This is an estimate, not an exact tokenizer. Real counts depend on the model's vocabulary — code, numbers and non-English text usually cost more tokens than this suggests. Treat the figure as a budget, not an invoice.
Three numbers worth knowing
Context
A long chat resends the whole history on every turn. The twentieth message costs far more than the first because of everything in front of it.
Output
Generated tokens are usually the expensive side. Ask for a short answer and you save more than by trimming the prompt.
Retries
Every retry is a fresh full-price call. A loop that quietly retries three times triples the bill for one answer.
Building something that calls a model at volume? See our AI automation work
Questions about tokens and cost
- Is my text sent anywhere?
- No. The count happens in this tab with plain JavaScript. There is no API call, so a contract, a medical note or an unreleased feature spec never leaves your machine.
- How accurate is the token count?
- It is an estimate with a ±15% band, not an exact tokenizer. Real counts come from a model's vocabulary file, which we do not ship. English prose lands close; code, numbers, emoji and Indic scripts usually use more tokens than the estimate shows.
- Why do I have to enter the price per million tokens?
- Because provider prices change and a hardcoded list goes stale quietly. Type the input and output rates from your provider's current pricing page and the calculator does the arithmetic. We never show a price we cannot stand behind.
- Why are input and output priced separately?
- Most providers charge more for generated tokens than for prompt tokens, often three to five times more. If you are budgeting a summarisation job, the document is input and the summary is output — price both.
- Why is the rupee figure only indicative?
- It converts at a stated constant rate on a stated date, not a live feed. Your card issuer or provider's invoice will use its own rate and may add a markup, so treat the INR number as a planning figure.
More tools: the full list.