LlamaNet
A free, open-source local LLM server that turns an idle Mac or GPU into a public, OpenAI-compatible inference node reachable over HTTPS, sharing spare compute with friends, apps or AI agents.
LlamaNet runs GGUF language models such as Mistral, Llama, Qwen, Phi, Gemma or DeepSeek-R1 locally through llama.cpp on Apple Silicon, an NVIDIA GPU or CPU, and exposes them behind an OpenAI-compatible chat completions endpoint. A single install script sets up an isolated virtual environment, installs llama-cpp-python, and joins the machine to LlamaNet's public discovery network, giving it a public HTTPS URL within about a minute through a Cloudflare tunnel with no manual configuration.
A local web-based Model Manager lets a user search for and download a GGUF model, then load and serve it with one click, and the network handles tunnels, request routing, discovery and load balancing across joined nodes so the user only has to run the model. It can also run on a rented GPU cloud instance rather than personal hardware.