Ollama runs open-source models locally — no API key, no data leaving your PC.

1. Install Ollama

Download the Windows installer from ollama.com and run it.

2. Run a model

In PowerShell:

ollama run llama3.1

First run downloads the model (a few GB), then drops you into a chat. Try a smaller one if RAM is tight:

ollama run phi3

ollama list        # models you have
ollama pull mistral
ollama rm phi3

Rule of thumb: ~8 GB RAM for a 7–8B model. Want it on a server instead? See VPS sizing on flowsmithy.