What it is
Models do not read letters or exact words, but tokens: fragments of text. A word can be one token or several. As a reference, about 750 tokens equal roughly 1000 words.
Why it matters to an agency
API pricing and the context limit are counted in tokens. Understanding them helps you estimate costs and stay under the context window limit.
Example
A long prompt with a lot of context uses more input tokens, which raises the cost of every call.
By the numbers
Rule of thumb: 1 token is about 4 characters or 3/4 of a word in English; in Spanish you get a bit less per token. About 750 tokens are close to 1000 words. API prices are quoted per million tokens, with output typically costing 3 to 5 times more than input. Unlike a word, a token does not respect boundaries: "automatización" can be split into 3 or 4 pieces.