Language models, like those used in AI, process and generate text using tokens, which are units of text smaller than words but larger than characters.