Word ↔ token estimator

A quick two-way converter between words and tokens. Handy for back-of-the-envelope budgeting from a word count (e.g. "a 500-word answer"). It uses a sensible average ratio — real counts vary by model and content, so calibrate with your own text below.

Calibrate with your own text (real tokenizer)
Words
Tokens (real)
Real tokens/word

Default ratio: ~1.33 tokens per word (≈ 0.75 words/token, ≈ 4 characters/token) for typical English. Code, other languages, numbers, and rare words run higher. The calibration box uses the real o200k_base tokenizer (exact for OpenAI; a close proxy for others).

More free token tools