Open-Source · Technologies
LM Studio
The easy way to see what local models can do.
LM Studio is a free desktop app for discovering, downloading and chatting with open models on your own machine. No terminal, no setup ceremony: pick a model, press download, start asking. It is the evaluation on-ramp for local AI, and we use it to prove what open models can do before anyone buys hardware.
MODEL EVALUATION · EXAMPLE DESK
TRIAL
MODEL CLASS
VERDICT
STATUS
Drafting-Test
General assistant
Fit for use
DONE
Coding-Test
Code model
Fit for use
DONE
Summary-Test
Small fast model
Fit for use
DONE
Reasoning-Test
Larger model
Marginal
REVIEW
vibes-only-eval
No test set
Unknown
RISK
Sample desk
done · review · risk
In plain terms
Do not buy the server before the model earns it.
LM Studio lets you test local AI on a laptop before committing to infrastructure. Here is what changes.
Without it
- Local AI debated in theory, decided on hype
- Hardware bought before any model proved itself
- Evaluation stuck behind terminal setup nobody has time for
- Sensitive test data sent to cloud tools just to try things
With it
- Models downloaded and tested over a lunch break
- Evidence from your real tasks before any spend
- A no code path that anyone on the team can drive
- Trials that stay entirely on the machine
What CG TECH can do with LM Studio
The work, broken into the parts that matter.
Download, chat, decide
A catalogue of open models, one click downloads and a familiar chat window, so evaluating a model takes minutes rather than a provisioning ticket.
model trials over a lunch break
Verdicts, not vibes
We turn trials into evidence: your real tasks as a test set, models compared like for like, and a clear read on what is fit for use.
evidence before hardware
From chat to code
LM Studio serves downloaded models through a local OpenAI compatible endpoint, so developers can prototype against local AI from the same desktop app.
prototypes against local models
Desktop first, server next
What proves itself in LM Studio graduates to a proper Ollama deployment on server hardware. Same models, deliberate path, no wasted spend.
a deliberate road to production
How an engagement runs
From curiosity to evidence, step by step.
01
Frame
We pick the tasks and the quality bar that matter to you.
02
Trial
Candidate models tested on your real work, on your machine.
03
Verdict
A plain read on what works, what does not and what it needs.
04
Path
A sizing and deployment plan for what earned its place.
Questions we hear a lot
Common questions about LM Studio
What is LM Studio?
LM Studio is a free desktop app for finding, downloading and chatting with open models on your own machine. There is no terminal and no setup ceremony: pick a model, download it and start asking.
What do we need to run it?
A reasonably modern computer. Smaller models run on ordinary laptops, and more memory or a GPU lets you run larger ones. The point is testing on hardware you already own.
Is it really free?
Yes, including for use at work. The costs are your time and, later, the infrastructure for anything you promote to production.
Does anything leave our machine?
Models are downloaded once, then chat and testing run locally. That makes it safe to trial with realistic material, applying the same care you would to any sensitive document handling.
LM Studio or Ollama?
Both, in order. LM Studio is the friendly desktop on-ramp for evaluating; Ollama is the server workhorse for production. Prove it in one, deploy it on the other.
What is it good for, and what is it not?
It is the quickest way to see what an open model can actually do before anyone buys hardware. It is not a server. For shared use and for applications, that is where Ollama comes in.
How large a model can our laptops run?
It comes down to memory more than anything else. Smaller models run comfortably on an ordinary work laptop, larger ones need a well specified machine. We test on your actual hardware rather than guessing.
Can we use it to test before committing to a build?
That is exactly how we use it. Run your real questions through a few open models on a laptop first. If the answers are good enough there, the case for a proper deployment writes itself.
Ready when you are
Wondering if local AI is good enough? Let us talk.
A discovery session maps your tasks, your options and your quick wins. You keep the plan either way.
What to expect
- A consultant replies within 4 business hours
- Session booked to understand your requirements
- We will provide you with a fixed price quote