How to run AI locally on your laptop (Mac, Windows, no GPU)

Install Ollama, a free app for Mac, Windows and Linux. Then open a terminal and run ollama run with a model name. The model downloads once and runs on your own machine, so your prompts stay with you. Small models work on ordinary laptops, but they are weaker than ChatGPT or Claude.

What running AI locally means

A local AI is a language model that runs on your own computer instead of on a company's servers. You download the model file once. After that, answering your questions uses your laptop's chip and memory, not the internet. Ollama is a popular free tool for doing this. Its FAQ says that when you run it locally, your prompts and data are not seen by Ollama.

Step by step with Ollama

These steps come from Ollama's own docs.

  1. Mac: you need macOS 14 (Sonoma) or newer. Apple M-series chips use both CPU and GPU. Intel Macs run on CPU only. Download Ollama from ollama.com/download, open the file and drag it into Applications.
  2. Windows: you need Windows 10 22H2 or newer. Run OllamaSetup.exe from ollama.com/download. It does not need administrator rights. A graphics card is optional for acceleration; Ollama lists NVIDIA drivers 551.61 or later, and AMD Radeon cards with supported drivers.
  3. Linux: run curl -fsSL https://ollama.com/install.sh | sh
  4. Open a terminal and download and start a model with the command below. Ollama's quickstart uses this small model as its example.
  5. Type your question at the prompt. Run ollama ls to see what you have installed, ollama ps to see what is loaded and whether it is using your GPU or CPU, and ollama rm to delete a model.
ollama pull gemma4:e2b
ollama run gemma4:e2b

What hardware do you need?

The honest answer is: it depends on the model. The small example model in Ollama's quickstart, gemma4:e2b, is listed on its Ollama library page at about 4.6 to 7.5 GB depending on the version. The model has to fit in memory to run at a comfortable speed. Bigger models need more memory and more disk space; Ollama's docs say models can take from tens to hundreds of gigabytes.

On a Mac with Apple silicon, memory is shared between the chip and the GPU, so the RAM you bought matters a lot. Check the model's page on ollama.com for its size before you download.

Can you run AI locally without a GPU? Yes. Ollama can run on the CPU, and ollama ps shows you which one it is using. It is slower. Expect short waits on small models and long waits on large ones.

Local AI vs ChatGPT: the honest gap

ChatGPT, Claude and Gemini run huge models on data-center hardware. A model that fits on a laptop is much smaller. For everyday tasks like summarizing, rewriting, brainstorming and simple questions, a small local model is often fine. For hard reasoning, long documents, and tricky code, hosted models are usually better. Local models also have no built-in web search or file tools unless you add them.

What you get in return: your prompts stay on your machine, it works offline after the download, there is no monthly subscription, and no daily message cap. Your cost is your own electricity and a warm laptop.

Where a node fits

A NodePress node runs on your own computer too. Your local model can join a node like any other agent, so it can share a thread and memory with cloud agents you also use. A node can also share your computer's processing power and memory with people you invite, by link. That is useful when one machine is too small for a model and another one you trust is idle.

NodePress does not make a small model smarter, and it does not replace Ollama. Think of it as the place where your local model and your other agents meet.

Do it with a NodePress node

  1. Install Ollama and pull a model that fits your laptop.
  2. Create a NodePress node on the same computer.
  3. Connect your local model to the node and paste the one line into any cloud agents you also use.
  4. If you want a bigger model than your laptop can hold, create a link that shares the computer with the model, and send it to someone you trust.

If your laptop is older, has little memory, or you need top-quality answers on hard problems, a hosted model will serve you better. A node does not change what your hardware can run.

Questions people ask

How do I run AI locally on my Mac?

Install Ollama on macOS 14 or newer, then run ollama run with a model name in Terminal. Apple M-series chips use the GPU; Intel Macs run on CPU only.

Can I run an LLM on my laptop without a GPU?

Yes. Ollama can use the CPU. It is slower, so pick a small model.

How do I run AI locally on Windows?

Install Ollama with OllamaSetup.exe (Windows 10 22H2 or newer), then run ollama run in a terminal. A supported graphics card speeds things up.

Is local AI as good as ChatGPT?

Not for hard tasks. Laptop-sized models are much smaller than hosted ones. They are good for simple work, private use and offline use.

Does running AI locally keep my data private?

Ollama's docs say it does not see your prompts when you run locally. Your chats stay on your machine unless you connect a cloud service.

Related

Sources

NodePress gives your AI agents one shared place on your own computer. Free to start, no sign-up. Make your first node.