On this page
Most people have never heard the word "token" outside of an arcade. Good news: the idea is simple.
Why the AI thinks in tokens
The AI doesn't read whole sentences the way you do. It breaks everything — your messages and its own replies — into these little chunks and works through them piece by piece. Tokens are simply the AI's bite-sized unit, the way minutes are the bite-sized unit of a phone plan.
That counter you see
The chat shows you a couple of numbers as you go. Here's what they're telling you:
- Context usage — roughly how full that "working memory" for this chat is getting. A long conversation uses more than a quick one.
- Context cost — what that particular message cost (in those tiny token units) against your balance. It's there so nothing's hidden from you.
The everyday way to picture it
Imagine the conversation is written on a whiteboard that the AI re-reads before every answer. Tokens are the words on the board. A short chat barely uses any space; a long, rambling one fills the board up. "Context usage" is how full the board is. "Cost" is the little tab for re-reading it that time.
Do you need to worry about any of this? Honestly, no. You can use the site happily and never glance at those numbers. They're there for the curious and the careful — not a test you have to pass. If a chat ever feels like it's getting forgetful or pricey, just start a new chat for a fresh, empty board.