Install Ollama on Windows

A quick, reliable setup: download the official installer, verify the CLI, pull your first model, and run locally — all in a few minutes.

⬇️ Download Ollama for Windows (Official)

Step‑by‑step

Download the installer

Open the official download page and get the Windows .exe.

Go to official download

Run the installer

Launch the .exe and follow prompts. The installer adds the ollama CLI and a background service. If a firewall dialog appears, allow local access.

Verify the CLI

Open Command Prompt or PowerShell and check that Ollama is available. If the command isn’t found, open a new terminal window.

ollama --version

Pull your first model

Download a starter model like Llama 3:

ollama pull llama3

Run locally

Start a local chat session with the model:

ollama run llama3

To verify the local API is up, you can also check the tags endpoint in a browser: http://localhost:11434/api/tags.

What’s next?

For better performance with NVIDIA/DirectML, see GPU Acceleration.
Explore more models in the Models Hub (Llama 3, Mistral, Qwen 2.5, Gemma 2, Phi‑4).
If something goes wrong, visit Troubleshooting.

Tips & common pitfalls

Firewall & port

Ollama runs a local server on localhost:11434. If prompted by Windows Firewall, allow local connections. Corporate firewalls may require exceptions.

Drivers & GPU

For GPU acceleration, keep your GPU drivers up to date. See GPU Acceleration for CUDA/DirectML notes.

Disk space

Models can take ~5–15 GB each depending on quantization and size. Ensure you have space on the drive where models are stored.

PATH not updated?

If ollama isn’t found, close and reopen your terminal. In rare cases, sign out and sign back in to refresh PATH.

Uninstall

Need to remove Ollama? Follow Uninstall to clean up the app and model files safely.

Browse models → Read the FAQ → How to update →

Community‑driven guide. Not affiliated with the official Ollama project.