The AI labs use each other's models constantly. It's also pragmatic: there are cases where one model can't do something but a different model can blow right through it.