The Max also has much more memory bandwidth. Especially for running local LLMs, that is the limiting factor much more than the number of GPU cores is.