Also available in: Português (Brasil)
Token counter
How many tokens does your text have? Paste it below and get an instant estimate of tokens, words, characters and approximate cost. Important: the exact number depends on each model's tokenizer, which is why we show a range instead of a single fixed value.
Rules of thumb
| Reference | Roughly equals |
|---|---|
| 1 token | about 4 characters in English |
| 1 token | about 0.75 word |
| 100 tokens | about 75 words |
| 1,000 tokens | roughly half a page of running text |
How this estimate is made
A token is the chunk of text the model actually reads: it can be a whole word, a piece of a word or a punctuation mark. The tool blends two classic approximations from OpenAI documentation: one based on character count (about 4 characters per token in English) and one based on word count (about 100 tokens for every 75 words). From the two it builds a range rather than a single number.
Why we show a range, not an exact number
There is no universal correct number. Every model uses its own tokenizer: the BPE family behind GPT splits text one way, the SentencePiece behind Llama another, so the same paragraph can produce different counts depending on the model. Portuguese, programming code, emojis and rare words tend to spend more tokens than plain English prose, so the real count can land above the top of the range in those cases. Treat the result as an honest estimate for planning cost and fitting the context window, not as the official count of a specific model.
The cost is approximate too. It multiplies the central token estimate by the price per million you type in, so it gives an order of magnitude. Real pricing varies by model and still separates input tokens from output tokens, which usually carry different rates.
Last updated: · Methodology and sources