Ollama is a free tool that downloads and runs large language models directly on your own computer, instead of sending your messages to a company's servers over the internet. Once a model is downloaded, it can run completely offline.

AI & ML — INTRODUCTION

What Is Ollama? Run LLMs on Your Own Computer, Explained

Every time you use a chatbot online, your message travels to a company’s servers, gets processed there, and the answer travels back. Ollama skips all of that — the entire model runs on your own computer instead.

The Quick Answer

Ollama is a free tool that downloads a large language model straight onto your own computer and runs it there — no internet connection required once it’s installed, no company seeing what you type, no API key, no monthly bill. You get a private, offline-capable version of the same kind of AI assistant you’d otherwise reach over the internet.

What the Experts Say

Ollama’s own project describes itself simply as a lightweight framework built for exactly this job — packaging up a model’s weights and settings into one bundle so it can be downloaded and run locally with minimal setup, rather than requiring a dedicated machine-learning environment to get started. See Ollama’s official documentation.

How It Actually Works

1
One command downloads a model. Ollama connects to a library of ready-to-use models — you type one command, and the model downloads to your computer.
2
Your own hardware does the thinking. Once downloaded, every response gets generated by your computer’s own CPU or GPU — no request ever leaves your machine.
3
Bigger models need more memory. A model’s size (measured in billions of parameters) roughly determines how much RAM your computer needs to run it comfortably.
4
It works offline. Once a model is downloaded, you can turn off your internet connection entirely and it will keep responding — nothing about generating a reply requires a live connection.

Pick a Model for Your Machine

Not every model fits on every computer — larger models need more memory. Select how much RAM your computer has and see which model sizes would run comfortably.

The reason these numbers are so much smaller than you’d expect for a model this size: quantization. And once several models fit your hardware, the next question is how to actually compare them.

See What Installing a Model Looks Like

Here’s a step-by-step walkthrough of what actually happens when someone installs and runs a model with Ollama for the first time. Click through it below.

👉 This is a scripted walkthrough for teaching purposes, not a live terminal — but every line matches what Ollama actually prints when you run this for real.

Fun Fact

Ollama isn’t limited to one company’s models — its library includes models from Meta, Google, Microsoft, and several independent AI labs, all installable with the exact same one-line command.

Test Yourself

What does Ollama actually do?

Once a model is downloaded through Ollama, what do you need to generate a response?

Ready to customize a model’s behavior and call it from your own code? Continue to the Intermediate level →

Not sure whether local AI is even the right fit for what you’re doing? See Cloud AI vs Local AI — which one to use, and when →