I disagree. If a model is good enough for your task/workload, then subsequent smarter models won't really help you.
This is an asinine argument. (Not to mention a complete goalpoast shift to what I was responding to)
One costs twice what the other does and produces identical output. Why, indeed.