Installation

Install Ollama on Windows step by step

1

Download OllamaSetup.exe

Go to the download page and get the official Windows installer. It is ~5 MB. Always download from the official source: ollama.com/download/windows.
2

Run the installer (no admin needed)

Double-click OllamaSetup.exe. A short wizard installs Ollama and starts the service automatically. You do not need administrator rights for a standard user install.

Windows SmartScreen may show a warning for unsigned installers. Click More info → Run anyway if you trust the source.
3

Verify Ollama is running

After install, open a new Command Prompt or PowerShell window and check the version:

cmd.exe — new window
C:\> ollama --version
ollama version 0.6.5

Always open a new terminal after installing. The terminal that was open before install will not see the updated PATH.
4

Allow through Windows Firewall

When Ollama first starts, Windows Firewall may prompt you. Click Allow access for private networks. This allows local apps (like a chat UI) to reach Ollama at localhost:11434. It does not expose anything to the internet.
5

Pull your first model

cmd.exe
# Download Llama 3 (~4.7 GB):
C:\> ollama pull llama3
pulling manifest...
pulling 8934d96d3f08... 100% ████████ 4.7 GB
success
# Or a smaller model (~2.2 GB):
C:\> ollama pull mistral

Browse all available models at Models Hub.
6

Run the model

cmd.exe
C:\> ollama run llama3
>>> Tell me about Windows 11
Windows 11 is Microsoft's latest desktop operating system...

Type /bye to exit the interactive session.

GPU detection

Does Ollama detect my GPU automatically?

Yes. Ollama automatically detects NVIDIA GPUs via CUDA and AMD GPUs via DirectML on Windows. You can verify which backend is active:

cmd.exe

C:\> ollama run llama3 --verbose

# Look for "gpu" in the stats output

total duration: 1.2s

load duration: 423ms

gpu layers: 32

If gpu layers shows 0, GPU is not being used. See GPU Acceleration guide for troubleshooting.

Tips & pitfalls

Common install issues

'ollama' is not recognized after installing

This means the PATH was not updated in the current terminal session. Close all open terminals and open a new one. If it still fails after a fresh terminal, see Not recognized fix.

Installer says "Windows protected your PC"

This is Windows SmartScreen for unsigned executables. Click More info then Run anyway. The official Ollama installer from ollama.com is safe.

GPU is not detected after install

For NVIDIA: update to the latest Game Ready or Studio driver (version 527+) and ensure CUDA is installed. For AMD: DirectML is built into Windows so it should work automatically. See GPU Acceleration.

Port 11434 is already in use

Another process is using the port Ollama needs. Find it with netstat -ano | findstr 11434 in Command Prompt. Stop the conflicting process or see Port 11434 guide.

Model download is very slow

Model files are 2–70 GB depending on the model. Download speed depends on your connection. You can pause and resume — Ollama resumes interrupted downloads automatically on the next ollama pull command.

Related guides