2 ms·

I think the best models will always require "the cloud" but there will be plenty of use cases where "good enough" local LLMs need to be on-device.