Token Size Calculator
Enter how many words or characters you have, the kind of text and the model's context window. The calculator estimates the tokens, the share of the window used, whether it fits and the cost at your price per million tokens.
- Words to tokens
- Context window fit
- Token cost
These are estimates. Each model's tokenizer splits text a little differently.
Language models read text as tokens, not words. A token is a piece of a word. In English prose there are about four characters or three quarters of a word in each token, so 1,000 words come to roughly 1,330 tokens. Code and many other languages need more tokens for the same text.
Words to tokens chart
English prose. Search for a length.
| Text | Words | Tokens |
|---|---|---|
| Tweet | 40 | 53 |
| 200 | 266 | |
| Blog post | 1200 | 1596 |
| Essay | 3000 | 3990 |
| Report | 10000 | 13300 |
| Short book | 50000 | 66500 |
| Novel | 100000 | 133000 |
About 1.33 tokens a word.
Words and context windows
English prose. Slide the length in thousands of words.
Each bar shows how much of a common window the text fills.
How to estimate tokens
Words to tokens. Step through it.
Count words
Or characters.
Count the reply too. A long prompt can leave too little room in the window for a long answer.
Prose, code or other languages
Token rates. Compare them.
Most efficient.
Many symbols.
Dense script.
Whitespace, symbols and rare words all use extra tokens.
Before you send
Prompt checks. Tick each item.
Tick each item as you go.
Trim repeated instructions and long examples to save tokens and cost.
What you get
Token planning. Each tile explains one result.
Estimate.
Share used.
Yes or chunks.
At 500 words.
At your price.
Input and output tokens are often priced differently. Use the price for the side you are estimating.
Frequently Asked Questions
No question matches that search.
How many tokens is a word?
About one and a third in English, so a hundred words are about a hundred and thirty-three tokens.
How many characters are in a token?
About four in English text.
What is a context window?
The most tokens a model can handle at once, counting both the prompt and the reply.
Why do code and other languages use more tokens?
Tokenizers learn common English pieces best. Symbols and other scripts break into more pieces.
How do I fit a long document in a small window?
Split it into chunks that each fit, with room left for the reply.
Are token counts the same for every model?
No. Each tokenizer differs, so counts can vary by ten to twenty per cent.
How is token cost worked out?
Tokens divided by a million, times the price per million.
Know your tokens first
Estimate tokens before you send a long prompt, so it fits the window and the cost holds no surprises.
More size calculators
Each one has its own calculator, worked examples and answers to common questions.