OnDeviceAI

30 vs 150 TPS: The local LLM debate exposing on-device AI's real problem

Reddit devs are fighting over what TPS counts as "fast enough" for local LLMs—one camp says 30, another insists 150. Underneath lies a real hardware t