In this homelab series video, I’m running the DeepSeek R1 model locally—specifically on my Minisforum MS-01—using Ollama plus the Open WebUI interface. DeepSeek R1 has been making waves for how efficient it is, and I wanted to show a practical “you can actually run this at home” setup, even if you’re not sitting on a monster GPU box. I don’t have a GPU in my homelab right now (yeah, I never thought I’d say that either), so I’m doing this CPU-only and focusing on getting you up and running rather than pretending it’ll feel like ChatGPT speed. I walk through spinning up a VM in my lab (you can skip that if you want to run it directly on your machine), then I use my usual automation flow—DNS entries, AWX/Ansible provisioning, Docker install, and Nginx reverse proxy—so I can hit the WebUI over HTTPS on my own domain. From there, I pull the DeepSeek R1 model from the Ollama library (I start with the default ~7B size), select it in Open WebUI, and run a quick test prompt. On CPU it’s slow (my example took about 17 minutes), but it’s still a solid way to experiment locally—and if you’ve got a GPU, it’s absolutely the move.