Their approach is different, they build ML tech that runs on-device, so whatever they developed has to be able to run efficiently on iPhone/iPad etc.

I don’t think we will “hear” about Apple using LLMs either way because they will no doubt call it something different like they always have.