Install Ollama and Get Your First Local AI Reply Quickly

I was sitting in my studio this morning, sipping coffee and staring at my Mac, thinking about all the ways local AI could simplify my work without relying on the cloud. I decided to take action right then, installing Ollama and setting up a model that could help me with everything from writing articles to managing my fashion brand's inventory.

First off, you need to understand that the model we'll install today isn't going to be the most advanced out there, it won't beat the top cloud models in complex reasoning tasks. But it's yours, running on your hardware, consuming your electricity, and requiring no internet connection. And for ninety percent of what I do every day, this local setup is more than sufficient.

To get started, we use Ollama. It's free, open-source, and the simplest way to run language models locally on a Mac. Other tools exist, some with more power, but in my experience, Ollama is the least frictional path from zero to a working model.

Here are the steps: install Ollama, pull a model, and ask it a question.

First up, installing Ollama. Open your browser and go to ollama.com. Click download and you'll get an application for Mac. Double-click the downloaded file and drag it into your Applications folder. There's no account creation or email entry needed, just drag and drop it in there. Once installed, you'll see a small llama icon near the clock in your menu bar. That means Ollama is running in the background.

Next, pulling a model. If you haven't used Terminal before, don't worry, it's easy. Press command space, type "Terminal," and hit enter. A window will open with a prompt. Type `ollama pull llama three point two`. This downloads about two gigabytes of model weights to your machine. On a normal internet connection, it takes one to five minutes. When the download finishes, you'll see the Terminal prompt again.

Finally, asking it a question. In Terminal, type `ollama run llama three point two` and press enter. After a couple of seconds, the prompt changes, allowing you to ask any question. For example, you can ask something like, "What is the best way to organise product photos for my clothing brand?" The model thinks for a moment and then starts answering, streaming the response word by word on your screen.

That's it, you have a local AI running on your Mac. If you have an Apple Silicon Mac with at least sixteen gigabytes of memory, llama three point two will run well. With more memory, like thirty-two or sixty-four gigabytes, you can run larger models that are smarter but slightly slower. You might try `ollama pull qwen two point five`, or for a top-end Mac Studio, `ollama pull llama three point one`.

Now, having an AI chatbot in your Terminal isn't very useful on its own. The real value comes from integrating it into your actual work.

One of the first things I do with every new local model is ask it to summarise something. For instance, if you have a long email or a meeting transcript, you can paste it into the prompt and ask for a three-point summary. Local models are surprisingly good at this task, even the smaller ones.

Another early use case is drafting content. If I need a first draft of an article, a reply to a customer, or a press release, I'll have the model generate a version that's rarely perfect but always saves me from staring at a blank page for ten minutes. It gives you something to build upon.

Lastly, I ask the model to explain things I don't understand. Whether it's a piece of code, legal terms in a contract, or medical jargon your doctor used, the local model provides a calm and accurate first explanation without trying to upsell anything.

Setting up Ollama and using these simple strategies will help you integrate AI into your work seamlessly and efficiently. The best part is that it's all running locally on your machine, giving you full control over your data and processes.