Self-hosted AI on a VPS
Local LLMs, image generation, AI agents — running on hardware you control. We cover what each tool actually needs, what works on CPU-only servers, and where a GPU starts to matter.
Tutorials
About this silo
"Self-hosted AI" used to mean "you need an H100." It no longer does. Quantized 7B and 14B models give surprisingly usable answers on a $20-30/mo CPU-only VPS, and the surrounding tools (Ollama, Open WebUI, LocalAI) have matured enough to be production-deployable. These guides reflect what actually works in 2026, not what was true in 2023.