Two setups. Picking the wrong one for your life is the most common mistake in this hobby. More common than picking the wrong model. - Setup one, everything on the phone: PocketPal AI, download a GGUF over wifi, chat fully offline. For travelers and the privacy-minded. Limits: small models, battery drain, downloads you do not want to do twice. - Setup two, phone as a window: Ollama, LM Studio, or Jan on a home computer, phone connects over the network. Bigger models, shared use. Limits: an always-on computer. - Start with setup one. Twenty minutes, free, teaches you what local models feel like. Move to setup two when you can name the phone's limits. Want it running without the setup? PrivateLLM deploy puts a private LLM on your AWS for $50 plus usage.