Count your text
| Model family | Estimated tokens | Basis |
|---|---|---|
| OpenAI (GPT) | 0 | ≈ 4 characters or 0.75 words per token (OpenAI docs) |
| Google Gemini | 0 | ≈ 4 characters per token; 100 tokens ≈ 60–80 words (Google docs) |
| Anthropic Claude | 0 | No published ratio — use Anthropic’s count_tokens API for exact numbers |
Paste text to see how much of a 1M-token context window it uses
What this tool measures
AI models don’t read words the way we do. They split text into tokens — chunks that can be a whole word, part of a word, a space or a punctuation mark. Everything about using AI is measured in tokens: how much text fits in a chatbot’s memory (its context window), how long an answer can be, and how much an API request costs. This counter gives you three kinds of numbers at once: plain text statistics (words, characters, sentences, paragraphs and reading time), token estimates for the major model families, and how much of a one-million-token context window your text would occupy.
How the estimates are calculated
We only use ratios that the providers themselves publish. OpenAI’s documentation gives a rough rule of thumb that one token is about four characters or 0.75 words of English text. Google’s Gemini documentation says a token is about four characters and that 100 tokens equal about 60–80 English words. Because the character-based and word-based rules don’t always agree, we show a range from the lower to the higher of the two.
Anthropic doesn’t publish a fixed ratio for Claude. Its token-counting documentation says that the same text produces roughly 30% more tokens on newer Claude models than on its earlier models, and recommends counting tokens with its API for each model. So instead of inventing a ratio, we show the OpenAI-style estimate as a rough minimum and point you to the official tool. If you need an exact count for billing, always use the provider’s tokenizer or counting endpoint.
How to use it
- Paste your prompt, document or draft into the box. The numbers update as you type.
- Read the token range for the model family you plan to use.
- Check the context bar. If a document uses a large share of the window, the model has less room for the conversation and its answer.
- For cost planning, copy the token estimate into the AI API cost calculator. Remember that the model’s answer (output tokens) is usually priced higher than your prompt (input tokens).
Worked example
Suppose you paste a 1,500-word report. Using OpenAI’s rule of 0.75 words per token, that’s about 2,000 tokens. Using Google’s range of 60–80 words per 100 tokens, it’s roughly 1,900–2,500 tokens. If you then ask for a 400-word summary, the answer adds another 530 or so tokens of output. Even with a small model, that’s a tiny fraction of a one-million-token context window — but if you plan to process 10,000 reports a month through an API, those tokens add up, which is exactly what the cost calculator helps you see.
Why token counts differ between models
Each model family uses its own tokenizer — a fixed vocabulary of text pieces learned from training data. Common English words often get a single token, while rare words, names, code, emoji and many non-English scripts are split into several pieces. That’s why the same paragraph can have a different token count in different models, and why counts for languages other than English are usually higher. Treat any English-based estimate as a lower bound for other languages. Our guide What are tokens? explains this in more depth with examples.
Remove this box by adding your content here, or hide all notes with
"showEditorNotes": false in site.config.json.Privacy
All counting happens in your browser. The text you paste isn’t stored, logged or sent anywhere — not to us and not to any AI company. When you close the tab, it’s gone. Even so, it’s a good habit to remove personal information from documents before you paste them into any chatbot; our AI privacy guide explains why.
Frequently asked questions
How accurate is this token counter?
It’s an estimate based on rules of thumb the providers publish for English text. Real counts depend on each model’s tokenizer, the language and the kind of text (code and numbers often use more tokens). For exact counts, use the provider’s own counting tool.
Why doesn’t it show an exact Claude number?
Anthropic doesn’t publish a characters-per-token ratio. Its documentation says newer Claude models use a tokenizer that produces roughly 30% more tokens for the same text than its earlier models, so we show a minimum and point you to Anthropic’s token counting API.
How many words is 1,000 tokens?
For English, OpenAI’s rule of thumb (1 token ≈ 0.75 words) gives about 750 words. Google’s guidance for Gemini (100 tokens ≈ 60–80 words) gives roughly 600–800 words.
Is my text uploaded?
No. Counting happens in your browser with JavaScript. Nothing is sent to our server or any AI company.
Do other languages use more tokens?
Often, yes. Tokenizers are usually most efficient for English. Many other languages, and especially non-Latin scripts, need more tokens for the same meaning, so treat English-based estimates as a lower bound.
Sources
- OpenAI — Key concepts (tokens rule of thumb) (opens in a new tab)
- Google — Understand and count tokens (Gemini) (opens in a new tab)
- Anthropic — Token counting (opens in a new tab)
Facts, prices and features were checked against these sources on November 19, 2026. AI products change quickly; if you spot something out of date, please tell us.