All workshops
In Spanish

Your first model running on your laptop

You install LM Studio, download a small model, and talk to it from your own code: no key, no per-call cost, no sending your data to anyone.

Oct7

Wednesday, October 7, 2026

05:30 PM a 07:00 PM GMT-6 · 90 min

Show in my time zone

Zoom

We email you the link when you hold your seat

Price

Free

Hold your seat

Leave your email to hold your seat.

Create your account and keep the material

The workshop material opens without an account. With one, you save your progress and keep all your workshops, their material and their recordings on your dashboard, always.

Create my free accountI already have an account
Sign in to enroll

What you'll do

Almost all the AI you use today runs in the cloud: fast, good, with a cost per call and a third party that sees your data. There are tasks where that doesn't matter, and tasks where it matters a lot: internal documents, customer data, processes that run a thousand times a day.

In this session we install LM Studio, download a small model that fits a normal laptop, and talk to it from a Python script, the same way you'd talk to a cloud provider, but on your machine. We measure live: how long it takes, how well it answers, where it gets things wrong more than a big model. That gives you a judgment you only get by trying it: which tasks can go local today and which can't yet.

We end by giving the local model tools, so you see how far a small model goes when it has to do more than answer.

  • LM Studio installed with a model running

    On your laptop, chosen according to the memory you have.

  • A script that talks to your local model

    Python, a few lines, and zero cost per call.

  • A measurement you made yourself

    Speed and quality against a cloud model, with your own prompts.

  • Judgment on when to go local

    Privacy, cost, volume: when it pays off and when it doesn't.

Who it's for

It stands on its own, nothing needed beforehand: all you need is Python and a laptop with at least 8 GB of RAM.

It's for you if…

  • You work with data you can't send to an external provider
  • You have a process with many calls and the cost is starting to add up
  • You want to know how far small models really are from the big ones
  • You want to try AI without creating accounts or paying per call

It's not for you if…

  • Your laptop has less than 8 GB of RAM (you'll get stuck on the download and no copy fixes that)
  • You want to fine-tune a model
  • You want an exhaustive comparison of open models

What to bring

Hardware matters here. Read this before signing up.

  • A laptop with at least 8 GB of RAM (16 GB is better). A Mac with an Apple chip, or Windows/Linux with enough RAM
  • About 5 GB of free disk for the model
  • Python 3.10 or newer. The base repo reaches you by email before the session
  • Install LM Studio and download the model before the session: Monday's email says which one to download; it takes a few minutes and keeps the room from waiting

How the time is split

The blocks, in this order.

  1. Installing LM Studio and choosing a model

    What fits in your memory and why. You download it from the app and it answers.

  2. Your code, talking to your laptop

    LM Studio answers like a cloud provider, but on your machine. One Python script and you're sending it prompts.

  3. Measuring

    The same prompts against a cloud model and against your laptop: time, quality, errors.

  4. Tools with a small model

    Tool calling with a local model: what it handles and where it breaks.

  5. When to go local

    The judgment on one page: privacy, cost, and volume.

Frequently asked questions

Do I need a GPU?

No. With 8 GB of RAM a small model runs on CPU, slower but enough for the session. A Mac with an Apple chip or a GPU runs noticeably better.

Do I need to know how to code?

Basic Python: running a script and changing a line. We give you the code ready to go; the session is about understanding what happens, not writing it from scratch.

Are local models as good as cloud ones?

In general, no; on narrow tasks, often yes. Measuring it with your own prompts is exactly the point of the session.

Is it recorded?

Yes. The recording goes out by email that same night to everyone who held a seat, together with the finished repo.

What if I get stuck?

The first minutes are for getting everyone's environment ready; we don't move on until the room is at the same step. And that night you get the recording and the finished repo.

What comes next?

If you want to keep going at your own pace, there are NIEVA's ecosystems. If you want to take it to production with support, we have the bootcamps; we cover it in the last minutes, no pressure.

An AI model running on your desk

No key, no cost per call, and the judgment to know when it's worth it.

Hold my seat