💬 Reddit 5d ago
24/7 Headless AI Server on Xiaomi 12 Pro (Snapdragon 8 Gen 1 + Ollama/Gemma4)
Developer converted Xiaomi 12 Pro smartphone into headless 24/7 LLM inference server running Gemma4 via Ollama with LineageOS, custom thermal management, and battery protection scripts. Uses ~9GB RAM for compute after stripping Android UI, with active cooling triggered at 45°C and charging capped at 80% for longevity. Demonstrates edge deployment of open-weights models on consumer mobile hardware.