A token is a small chunk of text, usually a piece of a word. Models read, write, count, and bill in tokens: your context window is measured in them, pricing follows input and output tokens, and speed and usage limits all trace back to token count.
A grocery checkout is the everyday version. The cashier doesn’t charge you by the letters printed on each package. They scan items, some tiny, some bundled, and the receipt adds up those units. Tokens are how the AI counts up your cart.
Why you care
Text isn’t free, and context is finite. Paste a ten-page PDF, a messy transcript, and five old email threads into one prompt, and the model spends tokens reading all of it before it writes a word back. A short, specific prompt with the right source material beats a giant pile of maybe-relevant text, saves money, and keeps you clear of a rate limit. Skip the exact math; clean inputs make AI work better.