AI models don’t count in words but in tokens. This tool estimates a text’s token count, its cost, and the share of context it occupies, and converts tokens, words, and pages.
How to use it
- Select the text’s language (French, English, or code), then paste your text or prompt.
- Read the estimated token count, its range, and the word and page counts.
- Optionally select a model to see the input cost and the percentage of the context window used.
- Use the converter to switch from tokens to words or pages, and vice versa.
The principles behind the calculation
- The text is analyzed directly in your browser: it is not sent anywhere.
- The estimation splits the text into words, numbers, line breaks, and punctuation marks. A word is worth its length divided by an average token length specific to the language; a number is worth one token per group of three digits.
- The displayed range goes from 15% below to 20% above the central estimate.
- The converter applies a words-per-token ratio specific to each language, and counts one page for 500 words.
Limitations to be aware of
Each model has its own tokenization: the estimate is accurate to within about 20%. For exact billing, use the publisher’s official tokenizer.
