Run it on your computer

Local is a setup choice, not a magic privacy shield.

A local runner such as LM Studio can keep prompts away from a hosted model service. Downloads, updates, extensions, backups, and the rest of your computer still matter. Cloud features and “improve the product” toggles can send text off-machine after all.

Lab

Rough memory teaching tool

Weight file 4.0 GB

Working memory

5.6 GB

About 1.4× the file, as a teaching estimate.

Fit

Many laptops

Context length and the app itself still add more.

This is not a compatibility promise. The app, context length, model design, and your hardware change the real requirement.

01
Install LM Studio (or a similar runner)
A graphical app. You do not start in the terminal.
02
Pick a small instruct model
Read the model card and the license. Open-weight is not automatically open-license.
03
Download a quantized file that fits
Quantization shrinks the weight file so a laptop can hold it. Quality drops as the file gets smaller.
04
Turn off remote fallbacks
Then test with made-up information before you type anything you care about.

Check

You downloaded an open-weight model. Are you automatically allowed to use it at work?

Choose an answer to see why.

Previous Next