Your First Private AI in 10 Minutes (No Terminal Needed)

By Jakub RusinowskiLast updated 6 min readBeginner

You can have a ChatGPT-style assistant running on your own computer in about ten minutes, without typing a single command. Once the model has downloaded, your conversations stay on your machine: nothing is sent to a server, no account is needed, and it keeps working with the Wi-Fi off.

Short answer: install the free Ollama app, choose a local model in its chat window, and send your first message. The first message downloads the model; every message after that runs on your computer.

What you need

  • A Windows or Mac computer. Windows 10 (version 22H2) or newer, or macOS 14 Sonoma or newer on an Apple M-series or Intel Mac. On Linux the app is not available; follow the terminal steps in Installing Ollama instead.
  • Free disk space for the model. The download size is in the table below.
  • An internet connection, once, for the app and the model download.
  • No account and no API key. Local models need neither.

Which model for your computer

Find the row that matches your computer and note the model name. That is the model you will pick in Step 2.

Your computerModel (Ollama)Download sizeReads imagesFitIn LM Studio, search
8 GB of RAM, no graphics card (an older laptop, a basic office PC)phi4-mini2.5 GBNoComfortablephi-4-mini
qwen3.5:4b3.3–4 GBYesComfortableQwen3.5 4B
16 GB laptop without a graphics card, or a 16 GB Mac (MacBook Air 16 GB, a 16 GB Windows laptop)gemma4:e4b6.6–9.5 GBYesTight. On a 16 GB Mac, close other apps first.gemma-4-e4b
8 GB graphics card (RTX 4060, RTX 3060 Ti)qwen3.5:4b3.3–4 GBYesComfortableQwen3.5 4B
16 GB graphics card, or a Mac with 24–32 GB (RTX 4070 Ti Super, Mac mini 24 GB, MacBook Pro 32 GB)gemma4:12b7.7–8 GBYesComfortablegemma-4-12b
24 GB graphics card, or a Mac with 36 GB (RTX 3090, RTX 4090, MacBook Pro 36 GB)gemma4:26b16–19 GBYesComfortablegemma-4-26b-a4b

Download sizes are from the Ollama library, checked 17 October 2026. The fit column is this site’s memory check, passed by every example machine in the row at the largest download size. For your exact computer, use the GPU & VRAM checker.

Not sure what is inside your computer? The GPU & VRAM checker detects it and shows every model that fits, and the Model Finder picks one for your goal.

Step 1: Install the Ollama app (about 3 minutes)

  1. Go to ollama.com/download and download Ollama for Windows or macOS.
  2. Windows: run the .exe file. It installs without admin rights. Mac: open the downloaded file, move Ollama to your Applications folder and open it. If macOS asks for permission, allow it in System Settings → Privacy & Security.
  3. The Ollama app opens with a chat window. On Windows it also sits in the system tray; on a Mac, in the menu bar.

If Windows SmartScreen or your antivirus stops the installer, or macOS says the app is damaged, see why SmartScreen blocks AI installers and Gatekeeper and local AI tools.

Step 2: Choose a local model and say hello (about 5 minutes)

  1. On first launch the app offers to sign in to use cloud models, or choose a local model. Choose a local model. You do not need to sign in.
  2. In the chat window, open the model picker in the message box and choose the model from your row of the table.
  3. Type a first message, for example "Hi! In two sentences, what can you help me with?", and send it.
  4. The first message starts the download. Wait for it to finish (this is the slow part, and it depends on your internet connection). Then the model answers.

That's it: you are chatting with an AI that runs on your own computer. Every message from now on is answered locally.

Step 3: Check that it is really private (1 minute)

  1. Turn off Wi-Fi (or unplug the network cable).
  2. Send another message. The model still answers, because it runs on your computer.
  3. Turn Wi-Fi back on when you are done.

One thing to watch: the model list in the app can also show cloud models. Those run on Ollama's servers, need an account, and your messages to them leave your computer. If you only want local AI, stick to the models you downloaded, or turn cloud off completely: Installing Ollama shows the one setting.

Prefer LM Studio?

LM Studio is the other good no-terminal option. It is free, including for work, though not open source. On a Mac it needs Apple Silicon and macOS 14 or newer.

  1. Download and install LM Studio from lmstudio.ai.
  2. Open the Discover tab and search for the name in the last column of the table above.
  3. Download the model, open the Chat tab, load the model and start typing.

You can install both, but run one at a time: they compete for the same memory. Ollama vs LM Studio vs Jan compares them in detail.

Five things to try first

  • Rewrite something: paste a rough email and ask "Make this clearer and friendlier, keep it short."
  • Summarise: paste a long article or meeting notes and ask for five bullet points.
  • Explain: "Explain compound interest as if I were twelve."
  • Brainstorm: "Ten names for a small bakery that sells sourdough."
  • Describe a photo: if your model reads images (see the table), drag a photo into the chat and ask what is in it.

Local models are smaller than the big cloud assistants. They are good at writing, summarising and explaining; for facts, check anything important, just as you would with any AI.

If something goes wrong

  • The download is slow or stops. It is a large file, so give it time and keep the app open. If it stops, quit and reopen Ollama, then send your message again to start the download again.
  • Answers come slowly. The model may be too big for your computer. Pick the model from the row above yours in the table, or check your machine with the GPU & VRAM checker.
  • The app says it cannot connect. Make sure Ollama is running: on Windows, look for its icon in the system tray; on a Mac, in the menu bar. Quit and reopen it if needed.
  • Something else. The troubleshooting guides cover the common errors on Windows, macOS and Linux.

FAQ

Do I need an account?

No. Local models in the Ollama app need no account and no API key. An account is only needed for Ollama's cloud models, which this guide does not use.

Is it free?

Yes. The Ollama app is free and open source, and the local models in this guide are free to download. The only running cost is the electricity your computer uses; the local vs cloud cost calculator works it out for your setup.

Does my data leave my computer?

Not with a local model. After the download, the model runs entirely on your computer, which is why it keeps answering with the Wi-Fi off. Cloud models are different: they run on Ollama's servers.

Does it work offline?

Yes, once the model has downloaded. You need the internet only for the first download and for app updates.

Will it run on my laptop?

Most laptops from the last few years run at least the smallest model in the table. For an exact answer, run the GPU & VRAM checker on your computer.

How do I keep it up to date?

On Windows and Mac, Ollama updates itself: when an update is ready, the app offers to restart to install it.

Next step

Now that you have a private assistant, find the model that best fits what you want to do with it: the Model Finder asks two questions and gives you one plan. Want to know what else stays private and why it matters? Read What is a private LLM?. For the terminal and the rest of Ollama's features, continue with Installing Ollama and the Complete Beginner's Guide.