LocalMind
On-device AI assistant for Android. No cloud. No signup.
Start new thread
shaharb

LocalMind - On-device AI assistant for Android. No cloud. No signup.

byā€¢
LocalMind runs an LLM (Qwen 2.5 / Gemma 2 / Phi-4 family) entirely on your phone via llama.cpp. No API calls, no telemetry, no signup. The 1.5B Q8 model fits in ~2GB RAM on a 2022-class phone. Conversations are kept in encrypted local storage; one tap to wipe. Works on airplane mode. Built for people who want AI help without their conversations sitting in someone else's cloud.

Add a comment

Replies

Best
shaharb
Maker
šŸ“Œ
Hi šŸ‘‹ Maker here. LocalMind is the on-device LLM for Android I built because every AI app I tried wanted to send my prompts to someone else's cloud. On LocalMind the model runs entirely on the phone via llama.cpp ā€” no API calls, no telemetry. Stack: GGUF-quantized models (Qwen 2.5 / Gemma 2 / Phi-4), CPU inference. The 1.5B Q8 fits in ~2GB RAM during inference on a 2022-class phone. Conversations stay in encrypted local storage; one tap to wipe. Things I'd love feedback on: ā€¢ Default model: 1.5B sweet spot, or should the smaller 0.5B be default? ā€¢ Any preset prompt template that would be killer built-in? ā€¢ Speed vs. quality tradeoff ā€” is on-device latency acceptable for you?