Hacker News
llama.cpp Gets a Consumer Front Door at llama.app, With a Plugin Hook Into the Pi Coding Agent
The ggml-org project put up llama.app, a plain-language landing page pitching llama.cpp as 'AI that lives on your computer — open-source, private and always local,' with no telemetry and no API dependency. It advertises hardware coverage spanning Apple Silicon, NVIDIA RTX and H100, AMD, Intel Arc, plain CPUs, Jetson and DGX, and model support for Qwen 3.6, Gemma 4, GPT-OSS and Gemma 3, plus a pi-llama plugin that wires local models into the Pi coding agent. The repo is cited at 123.5K stars. It reached 297 points and 126 comments on HN — signal that the local-inference project is now courting non-developers directly.
↳ Follow the thread