Because they have more capable LLMs and the hardware needed to run them.