You can run an LLM on a phone (tried it myself once, with llama.cpp), but even on the simplest model I could find it was doing maybe one word every few seconds while using up 100% of the CPU. The quality is terrible, and your battery wouldn’t last an hour.
deleted by creator