Most people first run into tokens when they hit a limit or open an API bill. The natural next question is simple: how many tokens is the thing I actually want to send? Here are realistic estimates for the kinds of text people work with every day.
The quick rule of thumb
For ordinary English, one token is about three quarters of a word, or roughly four characters. Flip that around and you get a handy shortcut: multiply your word count by about 1.33 to estimate tokens.
It’s an estimate, not a law. Short, common words often count as a single token. Long, rare, or technical words get split into pieces. We explain how that splitting works in How AI Models Actually Read Your Text.
Estimates for common text
| What you’re sending | Typical words | Approximate tokens |
|---|---|---|
| A text message | 25 | 35 |
| A short email | 150 | 200 |
| A single-spaced page | 500 | 670 |
| A typical blog post | 1,500 | 2,000 |
| A long report or research paper | 8,000 | 10,700 |
| An average novel | 90,000 | 120,000 |
These are ballpark figures for plain English prose. Your own text could land higher or lower.
Why your number might be higher
A few things push token counts up:
- Formatting. Bullet points, tables, line breaks, and markdown symbols all take tokens.
- Numbers and dates. Long numbers often get split into several tokens.
- Names and jargon. Product names, technical terms, and uncommon words tend to break into pieces.
- The model itself. Each provider uses its own tokenizer. Anthropic’s newer Claude models, for example, produce roughly 30% more tokens for the same text than earlier Claude models did.
Why this is worth knowing
Knowing roughly how big your text is helps with two decisions. First, will it fit? A novel at around 120,000 tokens fits comfortably in many current models, but not all of them. Second, what will it cost? Token count times the price per token is your bill, and that adds up fast if you’re processing documents at volume. Our token pricing guide walks through how that math works.
Check your own text
Rules of thumb are fine for a gut check, but the real number depends on your actual words and your chosen model. Paste your text into the calculator to see the exact token count and cost across 10 models side by side.
Leave a Reply