Comment by zmarty

I have been trying the small llama 2 models (7B and 13B) and they are now good enough for most of my NLP tasks. Plus, Qualcomm will support llama models on mobile devices: https://www.qualcomm.com/news/releases/2023/07/qualcomm-work... soon.

I think Apple will come up with some crazy hardware to run good quality LLMs.