Token
FoundationsCore AI Concepts
A token is the unit by which AI models read and generate text. Tokens are smaller than words: the standard rule of thumb is that one token equals roughly three-quarters of an English word, meaning 100 tokens equates to approximately 75 words. API pricing, context window limits and inference costs are all denominated in tokens.
In practice
Most enterprise teams only encounter this term when the bill arrives, by which point usage is already set. Put token volume into the business case at the start, because price per token multiplied by tokens per task multiplied by tasks per month is the whole of an AI running cost, and long documents pushed through a model repeatedly are where it escapes.