Yeah, seriously... a little more speed and 7b quantized LLMs would be almost usable. I guess that could be nice? :) But in practice, performance is mostly bein

Yeah, seriously... a little more speed and 7b quantized LLMs would be almost usable. I guess that could be nice? :) But in practice, performance is mostly being wasted on phones right now.