2 million tokens is about 1,473,684 words of English on GPT-5.6, or 2,947 pages of 500 words. On Claude 5.x models, which produce more tokens for the same text, it's about 1,017,544 words.
| Tokenizer | Words | Pages |
|---|---|---|
| GPT-5.6 | 1,473,684 | 2,947 |
| Claude 5.x | 1,017,544 | 2,035 |
| Claude Haiku 4.5 | 1,333,333 | 2,667 |
| Language | Words |
|---|---|
| Portuguese | 1,200,000 |
| Spanish | 1,254,237 |
| English | 1,473,684 |
| Code (words ≈ terms) | 1,100,000 |
Estimates from average characters per token. Real counts vary with vocabulary: technical text, names and numbers produce more tokens.
Tokenizers were trained on more English text, so words in other languages get split into more pieces.
The APIs return the real count in the usage field of every response; the token counter estimates it before you send.
Estimate based on public prices as of October 2026. Token counts vary with the text and the model; check the official pricing page before you decide.