Ollama
Run open-source models locally. One command, no coding.
Use it if
- You want to try AI without sending your text to anyone
- You have a reasonably modern machine with spare disk space
- You want to experiment without a subscription
Skip it if
- Your laptop is old or has under 8GB of memory
- You need the strongest possible model quality
- You never want to open a terminal at all
Ollama is the shortest path from “I wonder if I could run AI on my own laptop” to it actually working. It handles the download, the setup and the running. You type one command and you are talking to a model that lives on your machine.
How it works
You install it like any normal app. Then, in a terminal:
ollama run llama3The first time, it downloads the model, which is a few gigabytes and takes a while. Every time after that, it starts in seconds. Nothing you type leaves your computer.
What it is good at
Private drafting, summarising, rewriting, and anything you would rather not paste into somebody else’s service. It is also the easiest way to develop a feel for what smaller models can and cannot do, which is hard to learn from articles.
Where it falls short
A model running on a laptop is smaller than a model running in a data centre, and it shows on harder tasks. Reasoning is weaker, long documents are harder, and speed depends entirely on your hardware. On a recent Mac it feels instant; on an old Windows laptop it can be slow enough to be annoying.
Check your available memory before you start. Under 8GB and the experience will be poor enough to put you off the whole idea, which would be a shame.
The step-by-step guide walks through the first run properly.