Self-Hosting

Curated entries for self-hosting from the SMF Clearinghouse.

AMD-Based Self-Hosting

ROCm-compatible AMD GPU builds for open-source inference and training stacks.

Intel Arc GPU and Xeon CPU setups for OpenVINO, IPEX, and CPU-optimized inference.

The default OS for local LLMs: Ubuntu, Debian, Fedora, and Arch options with CUDA and ROCm support.

Apple Silicon local inference from entry-level MacBook Air to Mac Studio Ultra.

Apple Silicon native inference with MLX, llama.cpp Metal backends, and local agent tools.

CUDA-accelerated local AI rigs from a minimal RTX workstation up to multi-GPU servers.

Run local LLMs on Windows with WSL2, native CUDA, and tools like LM Studio and Ollama.