I would think that logic only works if opensource models like llama don't catch up? In other words the models don't become a commodity.