iPhone. Doable, not great. iOS is stricter about background activity and file access, so true on-device AI is a tighter squeeze: possible, not as frictionless as Android. - PocketPal AI ships on iOS too: offline chat with small quantized models. Wifi for the multi-gigabyte downloads. Patience. Modest expectations. - Most iPhone owners land on the second pattern: the model on a Mac or home server running Ollama or LM Studio, the iPhone as the chat window. - The tradeoff: bigger models, the phone stays cool, nothing touches a public cloud. But you need an always-on machine in the corner. - Travel a lot with no network? On-device. Mostly chat from home? Remote. Want it running without the setup? PrivateLLM deploy puts a private LLM on your AWS for $50 plus usage.