Token count estimator
A quick, vendor-neutral estimate of how many tokens a piece of text will cost you — without sending it anywhere.
- Characters
- 0
- Words
- 0
- Estimated tokens
- 0
- Non-ASCII chars
- 0
Estimate uses ~4 characters per token for Latin text and counts CJK and other non-ASCII characters as roughly one token each. Exact counts vary by tokenizer.
Know your token count? Price it across 30 models →
Frequently asked questions
How does this estimate tokens?
It uses the common rule of thumb that English text averages about four characters per token, and about 0.75 words per token. It also counts non-ASCII characters separately, since CJK text is much denser — often one to two characters per token.
When is the estimate wrong?
It is wrong on repetitive text (your estimate is high), on dense code and JSON with long identifiers (high), and on numbers and unusual punctuation (low). Treat it as a planning figure and confirm against your vendor's tokenizer when the bill matters.
Why not just call the real tokenizer?
Most exact tokenizers ship as large language-specific libraries. Running them in the browser would be slower and only work for one vendor's models. This tool is the fast, vendor-neutral approximation.
Does my text leave the browser?
No. The estimator is a few lines of JavaScript and runs locally. Nothing is uploaded, and nothing is stored.