Run AI on your own computer (free and private)

Going deeper

Part of: Understand AI leaderboards · Step 5 of 5

Facts checked October 22, 2026 against the official pages linked below.

You don’t need a subscription to use an AI chat assistant. A free app can download an AI model and run it on your own computer. When you use a model that runs on your computer, what you type doesn’t need to leave your device.

The trade-off is that your computer does all the work. It needs enough memory and disk space, and the models that fit on a home computer are usually smaller than the big online ones. New to the words? See open weights and parameters in the glossary.

Two free apps

Ollama. For Windows, macOS and Linux, from ollama.com/download. Its quickstart shows how to start your first chat.
LM Studio. A desktop app with a chat window, for Mac, Windows and Linux, from lmstudio.ai/download.

Both apps are free for running models on your own computer. Both companies also sell optional paid plans for cloud models, which run on their servers instead of yours (Ollama pricing, LM Studio pricing). You don’t need them for this guide. If privacy is the reason you’re here, pick a model that runs locally, not a cloud one.

What your computer needs

From each app’s own pages, checked October 22, 2026:

Ollama

Windows: Windows 10 22H2 or newer, Home or Pro. At least 4 GB of free space for the app itself, plus more for models, which Ollama says “can be tens to hundreds of GB in size” (Ollama on Windows).
Mac: macOS Sonoma (14) or newer. Apple M-series Macs use both the processor and the graphics chip; older Intel Macs use the processor only, so they’re slower (Ollama on macOS).
Linux: see the Ollama Linux page.

LM Studio (system requirements)

Mac: Apple Silicon (M1 to M4) with macOS 14.0 or newer. Intel Macs aren’t supported. 16 GB of memory (RAM) or more is recommended; on 8 GB Macs, stick to smaller models.
Windows: standard (x64) or ARM computers. On x64, the processor must support a feature called AVX2. At least 16 GB of RAM is recommended.
Linux: x64 and ARM64 are supported.

Why memory matters most

A model has to fit in your computer’s memory to run well. Bigger models need more memory and more disk space. If a model is too big, it runs very slowly or not at all. Start with a small model and only move up if your computer copes. Don’t buy new hardware just for this until you’ve tried a small model on what you already have.

Try it

Install Ollama or LM Studio from the official link above. Never use a download site, a pop-up or an ad.
Ollama: the quickstart gives a one-line command that downloads a small model and starts a chat. Type /bye to leave. LM Studio: pick a small model inside the app, download it, then start a chat.
Ask it something you’d normally ask an online assistant, and compare the answers.

What to expect

Answers can be slower than online assistants, especially on older computers.
Smaller models make more mistakes. Check anything important: see Check an AI answer before you trust it.
Model downloads are large. Use a connection without a data cap.
Local doesn’t mean careless. Keep your computer updated and only download models from inside the app you installed.

The short version

Local AI is free and keeps your words on your device, but it’s limited by your computer’s memory. Use the official installers, start with a small model, choose local rather than cloud models, and keep using free AI models safely. For more free tools, see Resources.