Ssorta? In the sense of “each term is randomly sampled from a probability distribution that depends on the current state” yes, but, they aren’t like an n-gram model (well, unless you actually make a large n-gram model, but that’s usually not what one is referring to when one says LLM).