Do you even need a GPU to start?
No. You can start today on the computer you already have.
This is the question that stops most people before they begin. They read a forum thread about someone’s expensive graphics card and quietly decide local AI is not for them. It is a false wall. You do not need to buy anything to take the first step.
The one thing that actually matters: memory
A model has to fit in memory to run. That is the real limit, not whether you own a fancy graphics card.
- On a laptop or phone, “memory” means your RAM - the same number you already know (8GB, 16GB, and so on).
- A graphics card (a GPU) has its own separate memory (VRAM), and it is faster at this work. But it is a nice-to-have, not a need-to-have.
So the honest rule is: more memory lets you run a bigger, smarter model. A GPU makes whatever you run go faster. Neither one is a gate to starting.
What to expect on what you have
| What you have | What you can run | What it feels like |
|---|---|---|
| Phone, or laptop with 8GB RAM | A small model (roughly 1-3 billion parameters) | Works. A little slow. Real. |
| Laptop with 16GB RAM | A comfortable everyday model (7-8 billion) | Smooth enough for chat |
| A GPU with 8GB+ VRAM | The same models, but noticeably faster | Snappy |
“Parameters” is just the model’s size - think of it like engine size. Bigger is more capable and needs more room. We will come back to it gently in a later post.
So where do you actually start?
On whatever is in front of you. Pick a small model, run it, and see it answer. The point of the first step is not power - it is proving to yourself that this runs on your machine, no permission required. The faster hardware can come later, if you ever even want it.
Ingredients: human + AI.