Wont be able to for long as smaller models keep improving performance with distillation.

Once apple and co successfully embed local LLMs into their devices and custom processors it’ll be quite hard politically to reverse course.