Comment by era37

LLMs are only going to improve by fragmenting them into specialized systems for low parameter high performance results. We’ve reached the point where models will get smaller and more compact