It's like how the "New Forest" is really old now: even small LLMs are (from what I've seen which isn't exhaustive) large compared to Markov language models.