🔒 100% private — runs in your browser, nothing is uploaded

Token Counter

Count characters and words and get a clearly labelled local token estimate — not any provider’s official count.

Language models read text in tokens — chunks that are often smaller than a word — so a word count alone does not tell you how much of a context window a passage will use. This counter gives you a transparent local estimate of that size, and it is an estimate in the strict sense: it is NOT any provider’s official count, and real tokenisers differ by model, vocabulary and language, so treat the number as a ballpark, never as a billing figure. The rule is fully documented so you can reproduce it by hand: split the text into runs of letters, runs of numbers and single punctuation characters; a letter or number run contributes one token per four characters, rounded up with a minimum of one, and each punctuation character contributes one. Worked example: “Hello world” is 11 characters and 2 words, and it estimates to 4 tokens — “Hello” is a five-letter run, ceil(5/4) = 2, and “world” is another 2 — a ratio of 2.00 tokens per word; adding punctuation, “Hello, world!” estimates to 6, because the comma and the exclamation mark add one each.

This is a local estimate, not an official count. The token figure below is an estimate produced by a simple rule on this page — it is NOT any provider’s official count, and real tokenisers differ by model, so the same text can tokenise differently in different systems. Use it for rough sizing and comparison, never for billing.

0
0
0
0

How the estimate works: the text is split into runs of letters, runs of numbers and single punctuation characters; each letter or number run contributes one token per four characters (rounded up, minimum one), and each punctuation character contributes one. Your text stays in page memory on your device only — it is never uploaded, never stored and never put into a link, and no price is shown because provider pricing changes and a stale price would mislead.

Quick answers

How many tokens is “Hello world”?

On this page’s documented estimate, 4 tokens: “Hello” is a five-letter run costing ceil(5/4) = 2 and “world” costs another 2, across 2 words and 11 characters — 2.00 tokens per word. “Hello, world!” estimates to 6 because the comma and exclamation mark add one token each.

That is a local estimate, not any provider’s official count — real tokenisers differ by model, so check the exact figure with your model’s own counter.

What is a token and why count tokens per word?

A token is the chunk size a language model actually reads, often smaller than a word. Tokens per word exposes token-heavy text at a glance: plain prose sits low, while code, URLs, tables and dense technical writing push the ratio up well before you hit a limit.

The word and character counts here are exact; only the token figure is an estimate, and it is labelled as one everywhere on the page.

How to use

  1. Type or paste your text into the box — the counts update as you type and nothing is sent anywhere.
  2. Read the four figures: characters (the full string length), words (runs separated by whitespace), the estimated tokens, and the tokens-per-word ratio.
  3. Use the estimate for rough sizing only: drafting a prompt, comparing two versions, or checking that a passage is in the right ballpark for a context window.
  4. For an exact figure, always use the tokeniser or counter supplied for the specific model you will actually call — real tokenisers differ by model, and this page’s estimate is deliberately simple and fully explained below.

About this tool

That ratio is the most useful number on the page. Dense technical prose, code, URLs and text full of short words all push tokens per word up, while plain flowing prose sits lower — so if a draft suddenly shows a much higher ratio than your usual writing, something in it (a pasted table, a long identifier, a block of code) is token-heavy and worth a look before you send it anywhere with a limit. For comparison, the plain Word Counter at /words/ gives words, sentences and reading time without any token estimate, and the two tools pair well: draft and polish there, size here. Deliberately missing from this page is any price: model pricing changes, differs by provider and would rot on a static page, and a confident stale price is worse than none — multiply the estimate by a current price from your provider’s own page if you need a cost. Your text itself never leaves the page.

Frequently asked questions

Is this the exact token count for my model?

No — and the page says so on purpose. This is a local estimate from a simple documented rule (letter and number runs at one token per four characters, rounded up, plus one per punctuation character). It is NOT any provider’s official count, and real tokenisers differ by model: the same text can produce different counts in different systems, especially for code, other languages and unusual words.

Why is the estimate different from a provider’s counter?

Real tokenisers are trained vocabularies: common words and word pieces often cost one token even when they are longer than four characters, while rare strings, names and non-English text can cost more than this rule predicts. The estimator here knows nothing about any vocabulary — that is the price of a rule you can verify by hand, and why it is labelled an estimate everywhere it appears.

How are words counted?

A word is a run of characters separated by whitespace. Leading and trailing spaces are ignored and any run of spaces, tabs or line breaks counts as one separator, so “ hello world again ” counts as exactly 3 words and an empty box counts as 0.

Why is there no cost estimate?

Because prices change and differ between providers and models, so any price printed here would quietly go stale and mislead. The page gives you the token estimate and the tokens-per-word ratio; multiply by a current price taken from your provider’s own pricing page for a cost you can trust. Your text is never uploaded, stored or put into a link in any case.

📦 Embed this tool on your site

<iframe src="https://verystep.com/embed/token-counter/" title="Token Counter" width="100%" height="520" loading="lazy" style="border:0"></iframe>

Free to use. The embed shows a small "Powered by Verystep" link.

First time embedding? Read the step-by-step guide →

More free tools