A token is a unit of text that an AI model processes. Models do not read words or letters directly. They split text into tokens, which can be a whole common word, part of a longer word, a single punctuation mark or a piece of code. A common rule of thumb for English is that a token is about three quarters of a word, but it varies by model and by language.
Code often uses more tokens than you would guess. Indentation, brackets, long variable names and unusual symbols all count. So do the files and logs the model reads, not just what you type. Text in languages other than English can also take more tokens for the same meaning.
Tokens matter for three practical reasons. Cost: API pricing is usually quoted per million tokens, with input (what you send) and output (what the model writes) priced separately and output typically costing more. Limits: subscriptions and APIs often cap how many tokens you can use in a period, which is why a heavy session can hit a usage limit. Context: the context window is measured in tokens, so a token-heavy conversation fills it sooner.
You can use fewer without getting worse results. Ask for the changed section of a file rather than the whole file back. Share the relevant part of a log. Start a new conversation instead of carrying a long one forward. Be specific in the first prompt, because a vague request often leads to several rounds of correction.
And yes, it is where the name of this store comes from.
In the wild
Hit my usage limit by lunch. Turns out I had it rewrite the whole 900-line file every time instead of just the function.
