I’m not optimistic. It’s the Wild West and comparing models for one’s specific use case is difficult, essentially impossible at scale.