Skip to main content

learn.howToCalculate

learn.whatIsHeading

A tokens to words calculator estimates the relationship between AI language model tokens and human-readable words. Tokenization splits text into subword units — most English words are 1–2 tokens.

Formule

words ≈ tokens × 0.75 (rough estimate; varies by tokenizer)
T
Tokens (tokens) — LLM token count
W
Words (words) — Approximate English word count

Guide étape par étape

  1. 1Rule of thumb: 1 token ≈ 0.75 words (or 4 characters)
  2. 21,000 tokens ≈ 750 words ≈ 3 pages A4
  3. 3Common words are usually 1 token; rare words may be 2–4 tokens
  4. 4GPT-4 context limit: 128K tokens ≈ 96,000 words

Exemples résolus

Entrée
1,000 words
Résultat
~1,333 tokens
Entrée
128,000 tokens (GPT-4 context)
Résultat
~96,000 words or ~384 A4 pages
Entrée
1 token
Résultat
~0.75 words or ~4 characters

Questions fréquentes

Why is the conversion approximate?

Different tokenizers (OpenAI, Anthropic, etc.) split text differently. BPE tokenization is probabilistic. A rough rule: 4 tokens ≈ 3 words.

What is a token?

A token is a subword unit. Common words = 1 token; rare words or punctuation = multiple tokens. Special tokens and formatting add overhead.

How accurate is the conversion?

For English, the 0.75 factor is a rough guideline. Expect ±10–20% variance depending on text complexity, language, and tokenizer.

Prêt à calculer ? Essayez la calculatrice gratuite Tokens to Words

Essayez-le vous-même →

Paramètres