Ollama on Windows: the setup guide
Ollama runs large language models directly on your PC — no account, no API key, no data leaving your machine. Here's how to get it working on Windows 10 or 11, which model to start with, and what to do when it doesn't run.
Last updated: September 2026
What Ollama actually is
Ollama is a small background service that downloads AI models and runs them on your own hardware. Once installed it sits quietly on localhost:11434, and any app on your PC can talk to it — including Clicky.
The trade-off against ChatGPT or Claude is straightforward: local models are smaller and less capable, but they are free forever, work with no internet connection, and nothing you type or show them ever leaves your computer.
Before you start: will it run on your PC?
The thing that matters is RAM. A model has to fit in memory to run, so pick by what you actually have:
| Your RAM | Start with | Download size |
|---|---|---|
| 8 GB | llama3.2:3b | ~2 GB |
| 16 GB | llama3.1:8b | ~4.7 GB |
| 32 GB+ | qwen2.5:14b | ~9 GB |
A dedicated GPU makes replies faster but is not required — Ollama falls back to your CPU. On 8 GB with no GPU, expect a readable pace rather than an instant one.
Install it
Download and run the installer
Get OllamaSetup.exe from ollama.com/download and run it. It is a normal Windows installer with nothing to configure, and it installs per-user, so you do not need administrator rights.
It lands in %LOCALAPPDATA%\Programs\Ollama and starts with Windows.
Pull your first model
Open PowerShell or Command Prompt and run:
ollama pull llama3.2:3bThis downloads about 2 GB. Nothing appears to happen for the first few seconds while it resolves the manifest — that is normal.
Check that it works
ollama list
ollama run llama3.2:3bThe first command lists what you have installed; the second drops you into a chat. Type a question, then /bye to exit.
When it does not work
"ollama is not recognized as a command"
The installer adds Ollama to your PATH, but terminals that were already open never pick that up. Close the terminal and open a new one. If it still fails, call it by full path:
& "$env:LOCALAPPDATA\Programs\Ollama\ollama.exe" listApps cannot reach Ollama
Ollama being installed and Ollama running are two different things. Check whether the server is actually up:
curl http://localhost:11434/api/tagsA JSON list of models means it is healthy. A connection error means the service is not running — launch Ollama from the Start menu, or run ollama serve.
A download stalls partway through
Pulls resume where they left off. Press Ctrl+C and run the same pull command again rather than starting from scratch.
%USERPROFILE%\.ollama\models and they add up quickly. Run ollama list occasionally and remove what you do not use with ollama rm <model>.Giving it eyes and a voice
Ollama on its own is a text box in a terminal. If you want something that sees your screen and answers out loud, that is what Clicky adds on top — hold Ctrl + Win, ask about whatever you are looking at, and it replies in speech while pointing at things on screen.
Clicky detects an existing Ollama install automatically, so if you have followed this guide there is nothing further to configure. See also: running AI fully offline on Windows.
Try it yourself — Clicky is free
A free AI assistant for Windows that reads your screen and answers out loud. No account, no API key, and it runs offline with Ollama.
↓ Download for Windows — freeWindows 10 / 11 · No sign-up