What does a sentence cost?

AI doesn't read letters or words — it reads tokens, small chunks of text. Tokens are the currency: they fill the context window, they show up on the bill, and they are the thinking budget. Here you can watch text break into them.

What you see

A text box with ready-made examples — below it, the text falls apart live into tokens, with a counter and a German/English comparison.

What it shows

AI's currency: tokens fill the context window, appear on the bill and are the thinking budget — and the same statement costs more in German than in English.

Try:

12 tokens · 69 characters

  1. Art
  2. ificial
  3. ·intelligence
  4. ·forget
  5. s
  6. ·the
  7. ·beginning
  8. ·of
  9. ·a
  10. ·long
  11. ·conversation
  12. .

· marks a space — it usually belongs to the word that follows. Raw byte values appear when a single character, like an emoji, is split across several tokens.

Same statement, German vs. English:

German 21 tokens “Künstliche Intelligenz vergisst den Anfang eines langen Gesprächs.”
English 12 tokens “Artificial intelligence forgets the beginning of a long conversation.”

More letters, far fewer tokens: English is the tokenizer's mother tongue, so the same statement costs three-quarters more in German here.

Shown with the open cl100k_base tokenizer (the vocabulary behind GPT-3.5/GPT-4), as of July 2026, computed right here in your browser. Claude, GPT & co. each use their own, similarly-working tokenizers — the exact boundaries differ, the principle doesn't.

Why it matters: how a window overflows is in the article on context rot; how model class, thinking depth and iterations set the price is in the three dials; and what happens when you economise on all three is in the ugly cat.

An unhandled error has occurred. Reload 🗙

Rejoining the server…

Rejoin failed… trying again in seconds.

Failed to rejoin.
Please retry or reload the page.

The session has been paused by the server.

Failed to resume the session.
Please retry or reload the page.