A language model does not process whole sentences or even whole words at once. Text is first broken into tokens, which are common chunks of characters, and each token is handled as a unit. A short word might be one token; a long or rare one, several.
This affects cost and limits, since models charge and reason per token, and it explains quirks like a model mishandling unusual spellings. Understanding that the model sees chunks, not letters, makes some of its behaviour far less mysterious.