Yes, and unlike the LLM they can iterate on a problem. When I multiply, I take it in chunks. Put the LLM into a loop, instruct it to keep track of where it is

Yes, and unlike the LLM they can iterate on a problem. When I multiply, I take it in chunks. Put the LLM into a loop, instruct it to keep track of where it is and have it solve a digit at a time. I bet it does just fine. See my other comment as to why I think that is.