Ollama is the simplest way to run a language model on your own computer — no cloud, no subscription, no ongoing cost. This page walks you through installing it, downloading models, and running your first local conversation.
On Windows, the easiest way is to download the installer from the Ollama website — it's a normal .exe file, just like any other program you'd install. Here's how:
Go to ollama.com and click the Download button. It detects your operating system and serves the Windows installer — a standard .exe file. (Or, if you prefer the command line, open PowerShell and run irm https://ollama.com/install.ps1 | iex.)
Double-click the downloaded file and let it run. It doesn't ask many questions — it installs Ollama, sets up a background service (you'll see an Ollama icon in your system tray), and makes the ollama command available. It takes under a minute on most machines.
Open a new PowerShell window and run:
You should see something like ollama version 0.x.x. If you get "command not found," close and reopen your terminal — the installer adds the command to your PATH, and a fresh window picks it up.
Ollama doesn't come with a model pre-installed — you choose which models to download. The command to fetch one is ollama pull, and the command to run one is ollama run. In fact, running a model automatically downloads it the first time if you haven't got it yet.
Let's pull a small, capable model that runs well on almost any machine:
This downloads a few gigabytes (the exact size depends on the model). The download happens once; after that, the model runs entirely offline with no further internet connection needed.
You'll see a >>> prompt. Type something and press Enter — it responds right there in the terminal. To leave the conversation, type /bye.
The full catalog of models is at ollama.com/library. Every model page shows the available sizes and their download sizes, so you can pick one that fits your machine. Here are the commands you'll use most:
Shows every model on your system, with its name, ID, size, and when it was modified.
Shows which model is loaded in memory and how much video memory it's using.
Shows a model's architecture, how many parameters (think "size") it has, and its capabilities.
Frees up disk space. Models are stored locally and stay until you remove them.
The number in a model's name (like 3B or 70B) tells you its size in billions of parameters. Here's the honest way to think about it:
"ollama is not recognized as a command"
Close and reopen your terminal. The installer adds the command to your PATH, but an open window won't see it until it's restarted.
Download stalls or fails
Check your internet connection and try again. If a partial download is the issue, run ollama pull again — it picks up where it left off, or re-downloads cleanly.
Out of memory when running a model
The model is too large for your machine's memory. Choose a smaller model, or look for a "quantized" (smaller) version of the same model in the library — they run with far less memory and little loss in quality.
Ollama's own website is an excellent place to explore further — it's the official source, so it's always current. Here's what's there and how to use it:
We keep this guide focused on getting you up and running. When you're ready to explore deeper — or hit a question we haven't covered — ollama.com is the best place to go. It's the source the models come from, so the information there is always up to date.
