These are more like Small Language Models since the amount of textual data from the past is extremely limited, and most of what's out there hasn't even been digitized.