Token Counter and Context Window Estimator: Will Your Text Fit?
Paste your text and set a characters-per-token ratio and a context window. You will see exact character and word counts, an estimate of the tokens, and whether the text and a reserve for the answer fit.
Last reviewed
Estimated tokens
38
3.23% of the window with the reserve; 123,866 tokens left
- Characters
- 152
- Without spaces
- 129
- Words
- 24
- Lines
- 1
- Words per token
- 0.63
- Fits in the window
- Yes
- More text that fits
- 495,464 characters
- Copies of this text that fit
- 3,260
Characters and words are exact. Tokens are an estimate: tokenizers differ between providers and between model versions, and code, numbers and other languages usually take more tokens per character than English prose. For an exact figure use your provider’s token counter.
Why this page estimates tokens instead of counting them
Language models read and write tokens, chunks of text that are often a word or part of one. Prompt limits and prices are measured in tokens, so you often want to know how many tokens a piece of text will use. An exact count needs the model provider’s own tokenizer, and tokenizers differ between providers and even between versions of one provider’s models. Anthropic’s token counting page, for example, says its count is an estimate that can differ by a small amount from actual use, and that models from Claude Opus 4.7 onward use a newer tokenizer that produces about 30 percent more tokens for the same text than earlier models.
Shipping a real tokenizer in a web page would add far more code than this site allows for a calculator, and it would still only be right for one provider. So this page does what it can do exactly and is honest about the rest:
- It counts characters (as Unicode code points, so an emoji is one), characters without spaces, words and lines exactly.
- It estimates tokens as characters divided by a characters-per-token ratio you choose. The box starts at 4, a rough figure for ordinary English prose, and you should change it to suit your tokenizer and your kind of text.
- It checks the estimate, plus tokens you want to keep for the answer, against a context window you enter.
Your text stays in your browser. It is not sent anywhere and is not put in the share link.
The calculation
- Estimated tokens = characters ÷ characters per token, rounded up.
- Window used = (estimated tokens + tokens reserved for the answer) ÷ context window.
- Tokens left = context window − estimated tokens − reserve. If it is negative, the text does not fit.
- More text that fits = tokens left × characters per token.
- Copies that fit = (context window − reserve) ÷ estimated tokens, rounded down.
Worked example
The sample text in the box is 152 characters and 24 words. At 4 characters per token that is 152 ÷ 4 = 38 tokens, which is 0.63 words per token.
Now a long document of 400,000 characters against a context window of 128,000 tokens, keeping 4,096 for the answer:
- Estimated tokens = 400,000 ÷ 4 = 100,000.
- Used with the reserve: 104,096 of 128,000, which is 81.33%.
- Tokens left = 128,000 − 100,000 − 4,096 = 23,904, room for about 95,616 more characters.
- Only 1 copy of the document fits in the window.
Characters and estimated tokens at 4 characters per token
| Characters | Estimated tokens |
|---|---|
| 100 | 25 |
| 1,000 | 250 |
| 4,000 | 1,000 |
| 40,000 | 10,000 |
| 400,000 | 100,000 |
Points to check
- Check the real count. Use your provider’s token counter when it matters, for example before sending a long prompt or choosing a plan. Treat this page as a quick guide.
- The ratio changes with the text. Code, numbers, markup and many languages other than English usually take more tokens per character than plain English prose.
- The ratio changes with the model. The same text can use quite different token counts on different tokenizers, so use a ratio measured for your model.
- Count the whole request. System prompts, earlier messages, tool definitions and attached files all count, not only the text you paste here.
- Leave room for the answer. The reserve box does this. A long prompt that fills the window leaves no space for a reply.
Related tools: the LLM API cost calculator, the AI subscription vs API calculator, the data storage converter and the percentage calculator.
Frequently asked questions
How many tokens is my text?
For a quick estimate, divide the number of characters by the characters-per-token ratio for your model. Ordinary English prose is often around 4 characters per token, but code, numbers and other languages usually take more tokens. For an exact figure use your provider’s token counter.
Why is the token count an estimate?
Exact counts need the provider’s own tokenizer, and tokenizers differ between providers and model versions. Anthropic’s documentation says its count is an estimate, and that newer models can produce about 30 percent more tokens for the same text than older ones.
Is my text sent anywhere?
No. Everything is worked out in your browser, and the text is not saved or placed in the share link. Only the settings, the ratio, the window and the reserve, can be shared. Close the page and the text is gone.
What is a context window?
It is the most tokens a model can take in at once, counting everything in the request and, on many models, the answer too. Enter the window for your model, and use the reserve box to keep tokens free for the reply.
How are characters counted?
As Unicode code points, so an emoji or an accented letter written as one character counts once. Spaces are counted in the characters total and left out of the without-spaces figure. Words are runs of text separated by spaces or line breaks.
What ratio should I use?
Start at 4 for English prose and lower it for code, numbers or languages that use many tokens per character. Better, count a sample of your own text with the provider’s counter, divide characters by tokens, and enter that ratio here.
Related calculators
- LLM API Cost CalculatorEnter the price per million input, output and cached tokens for up to three models, and your requests and token counts.
- AI Subscription vs API Break-Even CalculatorEnter the subscription price, how many requests each user makes and the API prices per million tokens.
- Data Storage ConverterConvert any data size.
- Percentage CalculatorPick the question you are actually asking.
- Bandwidth CalculatorChoose what to work out.
- Average CalculatorPaste a list of numbers to get the average, with the sum, count, smallest and largest.
- GPU Cloud Cost ComparisonEnter up to three quotes for GPU time, your GPUs, useful hours and utilisation.