setup 05reel 202026-08-13comment LOCAL
Run AI on a laptop with no graphics card
Ollama runs qwen3:4b on an 8GB laptop with no GPU, 100% CPU, fully offline after the first download.
You do not need a GPU. Without one, the model loads into normal system memory and runs on your processor. It is slower, and it works.
This is written for the machine most people actually have: 8GB of RAM, integrated graphics, no gaming card.
Install
Mac and Windows: download from ollama.com.
Linux, one command:
curl -fsSL https://ollama.com/install.sh | sh
Pick a model that fits
The download size is the floor, not the whole cost. Leave headroom for your context and everything else your laptop is doing.
| Model | Download | Where it fits |
|---|---|---|
qwen3:0.6b |
0.52 GB | anything, even a very old machine |
qwen3:1.7b |
1.4 GB | 8GB laptop, comfortable |
qwen3:4b |
2.5 GB | 8GB laptop, the sweet spot if nothing else is open |
Start it:
ollama run qwen3:4b
That is the entire setup. First run downloads the model, after that it works with the wifi off.
Check where it is actually running
ollama ps
Ollama's docs put it plainly: 100% CPU means the model was loaded entirely in system memory. That is the expected state on a machine with no GPU, not an error.
What this is good at
- Rewriting, summarising, and cleaning up text
- Drafting things you were going to write anyway
- Working offline, on a plane, with no account and no data leaving the laptop
- Anything you would feel weird pasting into someone else's server
What it will not do
It will not match the big cloud models. A 4B model is not GPT or Claude. It is a competent assistant, not a genius, and it will confidently get facts wrong.
It will be slow. Words arrive at reading speed rather than instantly. For a 4B on CPU that is usually fine for writing tasks and irritating for long conversations.
Long context eats RAM. The context window costs memory on top of the model. If it starts swapping, drop to the 1.7b or shorten what you paste in.
If you have more to work with
A machine with a real GPU, or an M-series Mac, can run much larger models the same way, same commands. The tradeoff only tightens when memory is scarce.
How much RAM has your laptop got?
Ask me in The Rabbit Hole and I'll tell you which one to start with.
It is the room where I keep the parts that don't fit on a page like this: the full run of numbers, what broke the first time, and the ones that looked good and then didn't work at all.
Two catches, up front, because that's the rule here. It's private, so you have to be inside to read anything. And it's new, so you'd be early rather than joining something busy.
if you run it, tell me
the rabbit hole
It's free to join, and the course library is free too. Ask there and I answer in the open, so the next person with the same question finds it. There's an optional paid Premium tier for weekly office hours.
join the rabbit holePost in there what you ran it on and what you got, or reply to the DM. I read every one.