Native Ollama on Windows — Local AI on Windows (Chapter 7)

Ollama publishes a native Windows installer at ollama.com/download. The Windows binary does not require WSL2—it runs as a native Windows service. This has practical advantages: files live in familiar Windows paths, no cross-filesystem navigation, and the service integrates with Windows Task Manager and Services.msc directly.

Install Ollama on Windows:

Download the installer from ollama.com/download
Run OllamaSetup.exe
The service starts automatically and listens on http://localhost:11434

Verify from PowerShell:

Invoke-WebRequest -Method POST -Body '{"model":"llama3.2:1b","prompt":"2+2=","stream":false}' http://localhost:11434/api/generate

Models are stored in %LOCALAPPDATA%\Ollama\models by default. Check disk usage:

dir "$env:LOCALAPPDATA\Ollama\models" -Recurse | Measure-Object -Property Length -Sum

The Windows installer creates a Windows service named "Ollama". Manage it with:

# Restart the service
Restart-Service Ollama
# Check status
Get-Service Ollama
# View logs
Get-Content "C:\Users\YOUR_USERNAME\AppData\Ollama\logs\ollama.exe.log" -Tail 50 -Wait

If you also run Ollama inside Docker (Chapter 6), both will try to use port 11434. The Windows Ollama will win by default. To use both simultaneously, change the Docker container's port mapping:

docker run -d --name ollama-docker -p 11435:11434 ollama/ollama:latest

Then access the Docker Ollama at http://localhost:11435.

Local verification checkpoint

Run the smallest example from this chapter in a local workspace and record the package version, runtime, data path, and observed output. If the result depends on model size, vector count, CPU/GPU backend, or available memory, note that constraint beside the exercise so the lesson remains reproducible.