What it is and what it is for
Language models do not read words, they read tokens, and every model has a token limit per conversation as well as a price per token. This tool shows the number of words, the number of characters and an estimate of tokens in real time as you type or paste text, without sending anything to a server: the calculation runs in your browser.
It helps you check whether a document fits in a model's context window, trim a prompt that has grown too long, and anticipate the cost before launching a large request to an API.
How to use it, step by step
- Paste or type your text in the box. The count updates on every keystroke.
- Look at the word count for a human reference of the length.
- Check the characters: they are counted as real characters, so an emoji counts as one.
- Use the token estimate as a rough guide, calculated at one token for every four characters.
- Compare the result with your model's limit and trim or split the text if you go over.
Worked example
If you paste the phrase "hello 😀 world", the tool shows 3 words, 13 characters and 4 estimated tokens. A 1,200-word English article is around 7,000 characters, which equals roughly 1,750 tokens by this rule, a comfortable size for almost any current model.
Frequently asked questions
Is the token estimate exact?
No. Each model uses its own tokenizer, so the real number may differ. The rule of one token per four characters is a handy approximation for budgeting.
Is my text saved or sent anywhere?
No. All the calculation happens in your browser and the text never leaves your device.
Why does non-English text use more tokens?
Tokenizers are trained mostly on English, so long words and accented characters in other languages are split into more pieces.
Does it handle emojis and line breaks correctly?
Yes. Characters are counted by Unicode code point and several consecutive line breaks count as a single word separator.