Expensive right now, but at the rate things are going, we'll probably see more efficient LLMs for things like mobile real soon.

I remember when voice to text was expensive, now we have on-device neural networks working.