Moreover, it's trivial to run several LLMs side-by-side for a while and measure the success of each, then migrate to the one that performs the best. And you can

Moreover, it's trivial to run several LLMs side-by-side for a while and measure the success of each, then migrate to the one that performs the best. And you can even migrate in-progress chats since all the context is passed on each call anyway.