AI Token Counter
Paste any text to estimate its tokens and see how much of each AI model’s context window it fills.
Count the tokens in your text
Your counts
How much of each context window does it use?
The context window is the maximum text a model can consider at once, including your prompt and its answer. Bars show your text against common window sizes.
What is a token?
A token is a small piece of text, often a word part rather than a whole word. The word “calculator” might be one token, while “tokenisation” splits into several. Models read and write tokens, providers bill by tokens, and context windows are measured in tokens, which is why this odd unit runs the whole AI world.
How this counter estimates tokens
Exact token counts need each provider’s own tokeniser, but two well-known rules of thumb get you close for English text: about 4 characters per token, and about 1.33 tokens per word. This counter averages the two estimates, which lands within a few percent of the real count for ordinary English prose. Code, URLs, numbers and non-English text can drift further, since they split into more tokens per word.
Quick reference: common sizes
- A short question: roughly 15 to 30 tokens
- A typical email: roughly 200 to 400 tokens
- A blog post of 1,500 words: roughly 2,000 tokens
- A 10-page document: roughly 6,000 to 8,000 tokens
- A full novel: roughly 100,000 to 150,000 tokens
Why context window usage matters
If your text plus the model’s answer exceeds the context window, the model either refuses, truncates the oldest text, or quietly forgets the beginning of the conversation. The bars above show how close you are to each common limit: 128K for GPT-4o class models, 200K for Claude class models, 400K for GPT-5 class models, and 1M for Gemini class models. Remember to leave room for the answer itself, a reply can be thousands of tokens on a long task.
What to do next with your count
Once you know the token count, pricing it is simple: providers charge per million tokens, split by input and output. Run your numbers through our AI Token Cost Calculator to see the per-request, daily and monthly cost on GPT, Claude, Gemini and DeepSeek models side by side. Our guide on how many tokens your prompt is also explains how to shrink a prompt without losing what matters.
Frequently asked questions
How many tokens are in a word?
For ordinary English, one word averages about 1.33 tokens, or put the other way, 100 tokens is roughly 75 words. Short common words are usually one token each, while long, rare or misspelled words split into two or more tokens.
How accurate is this token counter?
It is an estimate based on the standard character and word rules, typically within a few percent for English prose. Your provider’s own tokeniser gives the exact figure, and counts vary slightly between model families, so treat this number as a reliable guide rather than a billing figure.
How many characters are in a token?
About 4 characters per token in English text is the usual average. That average hides a lot: spaces and punctuation count, common words compress well, and strings like URLs, code and long numbers use far more tokens per character.
What happens if my text is bigger than the context window?
The model cannot use text beyond its window. Depending on the app, the excess is rejected, cut off, or the oldest messages are dropped, which is why chatbots sometimes forget the start of a long conversation. Split large documents into chunks, or summarise them first.
Does this tool send my text to a server?
No. The counting runs entirely in your browser with JavaScript. Nothing you paste is uploaded, stored or shared, so it is safe for drafts and private documents.
Why does my provider show a different count?
Each model family uses its own tokeniser with its own dictionary of text pieces, so GPT, Claude and Gemini counts for the same text differ by a few percent. Code and non-English languages differ the most. This counter averages the common rules to give a fair middle estimate.
How do I reduce the tokens in my prompt?
Cut filler instructions, remove duplicated examples, summarise long background documents, and drop old conversation turns. Then re-count here to confirm the saving, and price the difference in our AI Token Cost Calculator if you are on an API.
More tools you may need
Ready to price this text? Use the AI Token Cost Calculator to see what it costs per request on every major model. Building a website instead? Generate your llms.txt file in seconds.
