1. Installation & script safety
"Anything that suggests curl into bash just plain sketches me out." – walrus01
The discussion repeatedly warns against blind curl|bash installers and advocates reproducible, version‑controlled builds (e.g., NixOS, manual git clone → cmake → make). Users stress that trusting a script is “about the same level of security as blindly running a Windows .exe from the internet.”
2. Build simplicity & cross‑hardware performance
"the full set of llama.cpp binaries builds in under 5 minutes [...] on a literally ten year old dual xeon." – walrus01
Building from source is described as “not hard” and fast enough for even legacy CPUs. Community members highlight that the same binary runs on diverse hardware (Vulkan, SYCL, ROCm) and that pre‑built releases let you run models like Gemma 3 on a MacBook M3 Pro without lengthy compilation.
3. Attribution & legitimacy of the web front‑end
"llama.cpp is the real thing, ollama was a fork that remained inferior." – HelloUsername
Talk centers on clarifying that llama.app is merely a UI wrapper around the open‑source llama.cpp project and is not an official Meta product. Several users criticize Ollama for not giving proper credit, while others note that llama.cpp’s maintainers have “done a stellar job maintaining the quality while still being fast to implement new models.”