That question seems synonymous with "are those [current] behemoth LLMs close to optimal?" and it seems obvious that they are not. I wonder what ideas there are

That question seems synonymous with "are those [current] behemoth LLMs close to optimal?" and it seems obvious that they are not. I wonder what ideas there are about how one could estimate the optimal size.