llama.cpp

357 points · 159 comments on HN · read original →

Points and comments are a snapshot, not live.

llama.app is the official home for llama.cpp, supporting local AI agents on any hardware.

llama.cpp, via llama.app, can be paired with the Pi local coding agent (pi-llama plugin) for fully offline AI assistance: no config, no API keys, files stay local. It supports Apple Silicon, M Ultra, RTX 5090, CPU, Jetson, H100, MI300, RTX 4090, A100, M Pro, M Max, DGX Spark, T4, Radeon RX, B200, Intel Arc, and RTX 3090. Serve with `llama serve`, install the plugin, and run Pi.

What commenters are saying

Several commenters confirmed the site's legitimacy via links from the llama.cpp GitHub repo. The dominant sentiment is distrust of curl|bash installation methods, with many advocating for building from source (cmake -B build && cmake --build build --config Release) or using VMs/containers. One user reported struggles compiling llama.cpp with OpenVINO support for an Arc A770; others noted compilation is fast on modern CPUs. Some promoted NixOS for declarative, reproducible builds. A few argued curl|sh is comparable to trusting any package manager.