LM Studio on Linux: The Easiest Way To Run Local AI (No Cloud Needed)

Duration

26:55

Captions

1

Language

EN

Published

Dec 8, 2025

Description

Take control of your AI workflow with LM Studio on Ubuntu 24.04 a powerful, beginner-friendly tool that lets you download, run, and experiment with open-source AI models entirely on your own machine. In this video, I walk through installing LM Studio, exploring the interface, loading models, and enabling the local API so you can build your own AI tools without relying on the cloud. Whether you're a Linux user, a homelab builder, or you're simply curious about local AI, this guide gives you everything you need to get started. https://lmstudio.ai Rocky Linux Supported by CIQ: https://ciq.com/products/rocky-linux/ CompTIA Linux+ Certifcation Course https://youtu.be/qNxuTRCRjoQ Remember to Like, Share, and Subscribe if you enjoyed the video! Also, if you are interested in more Linux content, please consider becoming a channel member so I can continue to produce great content! ✔️RECOMMENDED LINUX BOOKLIST ------------------------------- Linux Pocket Guide: Essential Commands: https://amzn.to/3xGPvsK CompTIA Linux+ Certification All-in-One Exam Guide: Exam XK0-004 https://amzn.to/3uQ3wmh 101 Labs - CompTIA Linux+ https://amzn.to/3vtj7rb How Linux Works: What Every Superuser Should Know https://amzn.to/3vrLkOO Linux Bible https://amzn.to/3rwEkPH ✔️SOCIAL NETWORKS ------------------------------- KeepItTechie: https://keepittechie.com/ Facebook: https://www.facebook.com/KeepItTechie Twitter: https://twitter.com/keepittechie Instagram: https://www.instagram.com/keepittechie/ Discord: https://discord.gg/RjZWuyd -------------------------------- ✔️RECORDING EQUIPMENT ------------------------------- Insta360 4K Webcam - https://amzn.to/3RddfgZ Rode Procaster Microphone - https://amzn.to/42RSInF RØDE RØDECaster Duo - https://amzn.to/4ct1T1X Cloudlifter CL-1 Mic Activator - https://amzn.to/4ic7BXv Logitech LED Streaming Light - https://amzn.to/4j7Z8FT -------------------------------- 0:00 – Intro & Why Local AI Matters 0:40 – What Is LM Studio? Features & Platforms 2:30 – Installing LM Studio AppImage on Ubuntu 24.04 7:10 – First Launch, Model Download & Performance Check (htop) 9:25 – LM Studio Interface Tour (Chat, Models, Server, Settings) 18:30 – Using LM Studio’s Local API with curl 23:02 – Tips, Hardware Advice & Final Thoughts + Motivation

Captions (1)

00:00

What's up everybody? Welcome back to

00:01

Keep It Techie, where I help you learn

00:02

Linux and break into the tech field one

00:04

command at a time. I'm Josh and today

00:07

we're checking out a tool a lot of

00:09

people have been asking about, and

00:11

that's LM Studio. And if you've been

00:13

curious about running AI models locally

00:16

on your own hardware with no cloud, no

00:19

subscription, total privacy, then LM

00:22

Studio [music] is a great starting

00:24

point. Now, before we jump in, do me a

00:26

quick favor. Go on, hit that like button

00:28

and [music] subscribe if you want more

00:30

Linux tutorials, home lab projects, and

00:33

open source tools. Let's get into it.

00:39

All right, so I pulled up their website.

00:41

It's lmstudio.ai.

00:43

And LM Studio is basically a desktop

00:45

application that lets you run large

00:47

language models entirely on your

00:49

machine. It could be a laptop, it could

00:52

be a desktop, as long as you got the

00:53

hardware and it'll adjust based on the

00:55

hardware that you have. and [music] run

00:57

those AI models on your system. Now,

00:59

some standout features include a clean

01:02

UI for downloading and managing models,

01:06

a built-in chat GPT style interface,

01:09

also CPU and GPU acceleration, and then

01:12

local API compatibility, API support,

01:15

also works on Linux, Windows, and Mac

01:18

OS. So, if you have a Mac, you can run

01:20

this on your Mac. You can run this on

01:21

your Windows computer as well. Now the

01:23

big appeal is its simplicity. Instead of

01:26

juggling command line options or

01:29

configuring multiple tools, LM Studio

01:32

gives you a unified workspace. As you

01:35

can see, if we scroll down, that is the

01:37

standard across all platforms. Now,

01:40

here's why I think LM Studio is worth

01:42

checking out. For one, you got your

01:44

local only AI. Your data never leaves

01:47

your device. It's beginner friendly. You

01:50

don't need deep AI knowledge to get

01:52

started. Open AI compatibility API, easy

01:55

integration with scripts and apps. Also,

01:57

flexible model options. So, you got

02:00

Llama, Jimma, Mixurro, Pi, and more. And

02:04

this is awesome because it works great

02:06

in home labs because you can run it on

02:08

your desktop and even on a a VM or

02:11

server. And a quick note, LM Studio can

02:14

be used alongside Olama and there are

02:18

ways to bridge the two so they share

02:21

models, but we're not going [music] to

02:22

cover that in this video today. Just

02:25

know the option exists if you're running

02:27

Olama elsewhere in your home lab. Now,

02:30

let's go on and hop over [music] to our

02:32

virtual machine so I can walk you guys

02:34

through the process of getting this

02:35

thing set up. All right, so I'm logged

02:36

into my virtual machine. This is

02:37

Yubuntu24.04

02:39

4 and LM Studio ships as a [music] app

02:42

image for Linux and that's what we'll

02:44

install today. So, what you want to do

02:46

first is head over to LM Studios website

02:50

and let's give it a couple seconds, but

02:52

let's search for LM Studio. Just go to

02:54

the website right fast. I know [music]

02:56

it's a.ai. So, let's just go to it.

02:59

Search right fast. And what we want to

03:01

do is download the app image. And so,

03:04

that'll download into our downloads

03:06

directory. All right. So, we good.

03:07

That's our app image is stored in, like

03:10

I said, in our downloads directory. But

03:12

the first thing we need to do is open up

03:14

the terminal. That's one thing I don't

03:16

like about Ubuntu. I wish they would pin

03:19

the terminal here. So, let's add it to

03:21

favorites right fast. And we can zoom in

03:24

a little bit so you guys can see it a

03:25

little better. And first thing you want

03:27

to do is run [music] update. So, you go

03:29

pseudo apps update just to verify you

03:32

don't have any [music] updates for the

03:34

system, which I think this system has

03:35

been updated. Just make sure and then

03:38

there are a couple dependencies you want

03:40

to get installed [music] from one is the

03:42

fuse package and I'll show you guys that

03:44

it requires app images require fuse and

03:47

then I think wget is used to download

03:51

the models which I'm sure it is already

03:54

installed and I'm going install

03:55

something else right fast too. So, let's

03:57

go pseudo apps [music] install and let's

03:59

just install a lip fuse uh 2. That's the

04:03

right package name. And then let's just

04:05

verify get is on there, which [music] I

04:06

know it's on there. And then on this

04:08

server or on this system, I know for

04:10

some reason it don't come with htop. You

04:13

would think htop would be a default

04:15

application on here, but anyway, as you

04:17

can see, it's going to install HTOP and

04:19

Fuse [music] on the system, and that'll

04:21

allow us to run our app image. What's

04:23

up, y'all? If you've been watching my

04:25

channel for a minute, you already know I

04:26

stay talking about Linux. And if you're

04:29

looking for a solid, [music] reliable

04:30

enterprise Linux distro, let me put you

04:32

on to Rocky Linux. This is the go-to

04:35

replacement for SOS. And it's built for

04:38

the community by the community.

04:41

It's got everything you need for [music]

04:43

a stable and secure Linux experience,

04:45

whether you're running servers, home

04:47

labs, or enterprise workloads. And the

04:49

best part, it's backed by CIQ, making

04:52

sure it stays rock [music] solid for the

04:54

long haul. So, if you're tired of these

04:56

companies pulling the plug on your

04:58

[music] favorite dros, Rocky Linux is

05:00

where you need to be. And I've covered

05:01

Rocky Linux [music] before. And trust

05:03

me, it's worth checking out. So, head

05:05

over to rocky linux.org to learn more

05:07

[music] and get started. Keep it techy.

05:09

Peace. All right, so we good. We got

05:11

those two packages installed. Now, one

05:13

thing you can do, you can run it

05:15

downloads directory if you want to, but

05:16

what you have to do is go into

05:18

properties first. I'm going to show you

05:19

guys the guey way, but then go into

05:22

permissions. And then what you want to

05:24

do is go down to allow executing file as

05:27

a program. And that will change those

05:29

permissions or make it executable, which

05:32

I'm not going to do it that way. I'm

05:33

going just go into the terminal. I'm

05:35

going to do it that way in the terminal

05:36

because what I'm going to do is move it

05:37

and make cleaner. Put it in a better

05:39

location. I'll put it in the op

05:41

directory. That way it's a just a little

05:43

bit more cleaner. This will I don't know

05:45

just make the application a little bit

05:47

more cleaner for your system and I

05:49

recommend you guys follow this. So what

05:51

I'm going to do is type pseudo and we

05:53

have to spell that. Let's go move and

05:56

then [music] let's go under our

05:57

downloads directory and then the LM app

06:00

image. And what we're going to do is

06:01

move it under our app directory and

06:04

we're going to name it something

06:06

different. I'm going to name it LM

06:07

Studio app. And you got to make sure you

06:10

spell it right. So app with a capital A

06:12

and image with a capital I. Go press

06:15

enter. It'll move it over there. And as

06:17

you see it disappeared from my downloads

06:18

directory. Now you want to go and make

06:21

it executable over there. So we need to

06:23

use pseudo because that directory is not

06:25

owned by my user. So we have to use

06:29

pseudo. It's owned by root. So we have

06:30

to specify it to make it executable. So

06:33

let's go under our app directory. And

06:36

then we need to look for that lm studio

06:39

[music] app image. And then what we want

06:41

to do is link it. What I'm going to do

06:42

is link it to our user local bin. That

06:45

way we could just run the command.

06:48

That's what I'm saying. So let's go ln

06:50

as for link. Let's go s and then we're

06:54

going to link our ops lm studio image.

06:57

And we're going to link that to our user

06:59

directory. And then we want to go local

07:02

bin. And then we just want to name it LM

07:05

Studio. And that's good to go. All

07:08

right. So, with that being done, all we

07:11

have to do is launch the application by

07:13

typing LM Studio. It'll know where it is

07:16

because it's in that bin directory. And

07:18

so, it'll find it. Boom. So, it'll load

07:20

up. And this is the first time we've

07:22

opened it up. So, it's going to go

07:23

through a little setup. So, I'll walk

07:25

you guys through that right fast. And

07:26

boom. There we go. Local AI on your

07:28

computer. Get started. Now, when this

07:30

thing is first booting up on my virtual

07:32

machine, it's going to slow down a whole

07:34

lot. At least for me. But if you got

07:36

like a better computer, you know what

07:38

I'm saying? Faster computer, you should

07:40

be fine. Or a laptop or something that

07:42

has a pretty good specs, you should be

07:44

good. And that's why I wanted to install

07:46

Htop so we could look at what's going

07:48

on. As you see, it's maxing out our CPUs

07:51

and all that stuff. We can watch what's

07:52

going on while it's going through the

07:54

process. And this is going to ask you

07:56

what level of basically logs you want to

07:58

see or the output that you want to see.

08:00

You could do power user, you can do

08:02

user. It's basically saying I'm just

08:04

getting started with AI or I know what

08:05

I'm doing or show me everything. If you

08:07

want to be a developer or you want to

08:09

have the developer level, I'm going just

08:11

select the power user. That's fine. One

08:13

thing it's going to do is look for a

08:15

model that will work on your system. And

08:18

it'll also add LM Studio command line to

08:21

your path. So, that's another additional

08:23

setting that it will add to your system.

08:25

So, let's go on and download this model

08:27

that it recommends. It's only 2.5 gigs.

08:30

[music] And so, just let that go until

08:32

it finishes. You can explore the app. I

08:35

would I recommend you just wait for it

08:36

to finish cuz every time I've clicked

08:38

that when I've been playing around with

08:40

LM [music] Studio, it'll make the

08:43

download fail of the model. So, I don't

08:45

know if that's a bug or something, but

08:47

it has made it fail for me. I don't know

08:50

if other people have seen that or if

08:52

that's just a bug they need to work on

08:54

or something. I don't know. But whenever

08:56

I don't go in there, it'll download the

08:58

full model and then we're good to go.

09:01

But I'll be back when this finishes. All

09:02

right. And just my luck, that mug

09:04

failed. So, what we're going to have to

09:05

do is just start the download again.

09:08

It's fine. It'll continue from where it

09:10

stopped at. But this right here is just

09:11

showing you some of the new features in

09:13

this latest version. And let's go back

09:15

under here under our downloads. And this

09:18

will allow us to resume it. So, let's

09:20

just let it go. And I'll be back when it

09:22

finishes. All right. So, it finished

09:24

downloading. And what you can do is load

09:26

that model right away. I'm going to go

09:28

and close this, but you hit load model.

09:30

And what it's going to do is load it in

09:31

there as your main model. And then you

09:34

can start chatting. But let's go over

09:36

the interface right fast. So, obviously,

09:38

this is your chat window. Right here is

09:40

the developer window. This allows you to

09:43

share your models. It It allows you to

09:46

set up a server. So you can share it

09:48

over the network and it's basically a

09:50

API tab and you can enable LM Studio

09:54

locally and it's like a open AI

09:56

compatible API and right now as you can

09:59

see it's turned off by default but you

10:01

can set all this up. You go into here

10:03

that's your server ports. You can set up

10:04

whatever ports you want. You can serve

10:06

it on the local network allow per

10:08

request like for MCP servers just in

10:11

time model loading all that good stuff.

10:14

And there's also examples on how the

10:16

imports work. And then also the logs

10:19

down here. And then you can go through

10:20

it's a lot more information here. You

10:21

can go in the context interface. This

10:23

gives you custom fields settings like

10:26

the temperature sampling. What else?

10:29

Structured output. You can do JSON

10:31

scheme. Let's see what else is on here.

10:34

Speculative decoding if you [music] need

10:36

it. And then let's see. Load. So, this

10:39

just goes through and you can make

10:40

adjustments under here for the load,

10:42

especially if you have a GPU or

10:44

something if you want to offload certain

10:46

things to a GPU. All right, so that's

10:47

enough about the server. Now, under

10:50

here, this shows you all your models you

10:52

have installed and as well as the

10:54

location where those models are stored.

10:56

So, this is our models directory. So,

10:57

under my home directory, there is a

10:59

hidden folder called LM Studio and that

11:01

is where my models are stored. [music]

11:03

And right now, we have this one model.

11:05

That's the only model we have currently.

11:07

And then it would click under here under

11:09

the search or the discover. This will

11:12

open up these settings and this will

11:13

[music] allow you to search for other

11:15

models if you want to get another one

11:17

installed. [music]

11:18

And there is a status pick. You can

11:20

refresh that. You can do best match.

11:22

[music] Uh and like I said, this thing

11:24

will show you based on what you run on

11:26

your system. It won't show you like big

11:29

models. Now, you could change this where

11:31

it'll show it'll show any and

11:33

everything, but you may not be able to

11:35

run it on your system. You know what I'm

11:36

saying? Because it may be like a huge

11:39

model that you need more RAM for. And as

11:42

you can see down here, it says based

11:43

[music] on a calculated device memory of

11:46

7.76 GB. So I have 8 gigs on this

11:49

server, which is equivalent to 7.76 GB

11:53

as you can see under my hype. That's why

11:55

I open up HTtop. So that's the max I can

11:57

use. It's calculating that and only

11:59

showing us models based on the amount of

12:02

RAM I have. And like I said, you can

12:04

look at some some of the big ones if you

12:06

want to, but it'll adjust based on what

12:08

you have on your system. [music] So,

12:10

just be wary of that. And also, let's go

12:12

under runtime. This will show you your

12:15

extensions, the packs, and everything.

12:17

Let's say like for Harmony. It's a chat

12:20

history, render, and porcing. [music] It

12:22

has a fix for this. You need to let's

12:26

see upgrade [music] Python environment.

12:27

This fixes a bug for the Harmony server.

12:29

So, let's go and fix that. We could just

12:31

run that. download the patch for [music]

12:34

it. That's open AI Harmony. That'll get

12:37

that dates on our system. [music] So,

12:39

update that for us. And we also have

12:42

some other extension packs in here. It

12:44

says error surveying hardware. Yeah,

12:47

that's because we don't have CUDA on

12:49

here, [music] which is for a video card.

12:50

We don't have a video cord on here, so

12:52

can't put CUDA on there. And let's see,

12:54

Vulcan. I'm not sure what that's for.

12:56

Yeah, GPU [music] required. So, we don't

12:58

have a GPU. So, that's why that's not

13:00

installed as well. You can go into here

13:02

and just look at what's all compatible

13:04

[music] with your system. We don't have

13:05

any of that stuff. So, we can't install

13:07

any of the other extensions or

13:09

frameworks and all that stuff. So, go

13:10

under our hardware and this breaks down

13:12

the hardware. You got offload KV cache

13:15

the GPU memory if we had it. Let's see.

13:17

CPU compatible, but as you can see,

13:19

yeah, it says zero GPUs detected. Let's

13:22

see. Memory capacity, it breaks that

13:23

down. Guard rails. Now, you can modify

13:25

your guard rails. You got balance,

13:27

relax, off, not recommended. custom.

13:31

That's your guard rails right there. So,

13:33

you want to make sure you pay attention

13:35

to that and don't go off the reservoir

13:38

with your gore rails. See, [music] you

13:40

can look at it right there. It says

13:41

loading models beyond system resource

13:43

limits may cause system instability or

13:46

freezing. So, if go outside them gore

13:49

rails, you know what I'm saying? You

13:50

could like kind of mess up your system.

13:51

You know what I'm saying? Where it

13:53

freezes up and you have to restart your

13:54

system or something like that or

13:56

whatever. But anyway, let's go under

13:57

settings. I just want to show you guys

13:58

some of the other settings right there.

14:00

You got general settings. Let's say you

14:02

want to stay on the stable version or

14:03

you want to go to the beta version. I

14:04

recommend you stay on the stable

14:06

version. That way you don't run into any

14:07

issues. You can do check for updates.

14:09

There we go. And then show side button

14:11

labels. You can put the labels up there.

14:14

I like to add that. You can also change

14:15

the colors. You see how the colors

14:17

changed over there. Presets. So show

14:19

configuration dialogue when committing

14:22

new fields to the presets. Auto updates.

14:24

Auto deletes. use LM Studio HuggyFace

14:27

proxy. So that's a proxy open download

14:30

pane when starting a new model download.

14:33

So that's what that is right there. When

14:35

we download it, it popped that [music]

14:36

open so we can see the downloads. And

14:38

this is the user interface complexity

14:40

level which we selected in the beginning

14:42

with the different levels. So user,

14:44

power user, developer. You can go in and

14:46

change that after the fact. And then you

14:47

can modify how you want to see the

14:49

models. You can look at the full name or

14:51

not, the color theme. You can specify

14:53

what you want. I'm going leave it on

14:54

auto cuz it's fine. Language, you can

14:57

modify that. Model defaults, you specify

15:00

that. Model maximum, this is for the

15:02

guard rails for loading the model. So,

15:05

beyond system resource limits, you just

15:07

want to keep those guard rails on there.

15:09

I recommend you keep those guard rails

15:11

on. And then you can reset all this if

15:12

you go in and mess around with it. But

15:14

just show you some more of the options.

15:16

So, under our chat, we got our chat

15:17

options. So, you can go in here, make

15:19

changes under there as well, like a few

15:22

changes. And then we got a developer.

15:24

This will show debugging [music]

15:26

information. Enable model load

15:27

configuration support and extension

15:29

packs. Let's see on demand loading all

15:32

that [music] good stuff. And the

15:33

integration right now we don't have any

15:35

integration. And then right here, this

15:37

will take you to the LM Studio hub. This

15:40

will allow you to go to the

15:41

documentation, all that stuff. And then

15:42

you can log in to LM Studio Hub. That

15:44

way you can get some more information.

15:46

You can join the organization, create an

15:48

account to publish projects, all that

15:51

good stuff. So, that's pretty much it

15:52

under the settings. I just wanted to at

15:54

least show you guys that. And that's

15:56

pretty much it on the interface. [music]

15:58

I mean, you can go up here and look at

15:59

the menu options. So, we got file,

16:01

that's quits, and actually it doesn't

16:03

quit. Let me show you guys that right

16:05

fast. So, if we hit quits, it's actually

16:07

not stopped. As you can see, the

16:09

terminal is still running back here in

16:10

the background. It's still running.

16:12

That's because it is in [music] the

16:14

tray. So, we can open it back up. It'll

16:16

open it back up. And we are back to it.

16:18

So, as you can see, runs in the

16:20

background. It'll stay running in the

16:22

background if you don't close it or if

16:24

you don't quit it. You [music] have to

16:25

quit it up here. You have to right click

16:26

on it and hit quit LM Studio to quit the

16:29

application. Just so you guys know, you

16:31

can undo all the normal stuff under edit

16:34

view. You [music] can change the view of

16:36

it. The windows zoom minimize close and

16:40

then under help that'll take you to the

16:41

technical documentation and the [music]

16:43

LM Studio blog and the website. And down

16:45

here you can see what's going on with

16:47

your system. So, the amount of RAM and

16:49

the CPU usage at the time and then your

16:51

account if you're logged into your

16:53

account. And then this will bring up

16:54

your settings as well. It'll [music] go

16:56

back into the settings for you. Now,

16:57

let's go back and play around with the

16:58

chat. Like I said, it's running on a

17:00

virtual machine, so this thing is slow,

17:02

but I'm going to go down and just run, I

17:05

don't know, just run something. Let's

17:07

see. [music] Explain LM Studio. Simple

17:11

terms. Let's see what it Let's see what

17:14

it does when I say that. Boom. And it

17:16

should give us some feedback here. It's

17:18

going to think, okay, the user wants me

17:19

to explain LM Studio in simple terms.

17:22

First, I need to make sure I understand

17:23

what LM Studio is. Wait, I'm not

17:26

entirely sure. See, and that's one thing

17:28

about the these models. They're offline.

17:31

So, they don't have all the context. And

17:34

so, some of the answers are not going to

17:37

be good. You have to connect this thing

17:39

to the internet or I'm not sure how to

17:41

connect to the internet. I haven't tried

17:43

to do that yet. I know how to connect

17:45

thlama to the internet. That way you can

17:48

get up toate information and it should

17:50

be the same way in here. Some kind of

17:51

way you can open it up to where you can

17:53

connect to the internet as well. I just

17:55

haven't done it. But if you look here,

17:57

this will show you what's going on with

17:58

your system CPU and your memory usage

18:01

and swap. It's going through. But that's

18:03

all running on your system while this

18:05

model is running in the background. Just

18:07

wanted you guys to see what's going on.

18:09

But as you can see, it's writing out

18:11

some information that we asked for. And

18:14

it even gives you some information down

18:16

here. So 3.52 tokens per second, 921

18:19

tokens, 2.53 seconds for to first token

18:24

stop reason token file and it thought

18:27

for about 2 minutes and 59 seconds to

18:29

come up with this results. Right. So

18:32

let's quickly try another model. Let me

18:34

download one right fast. Actually, let's

18:36

go to discover right fast and let's look

18:40

for I want to go to yeah Pi Mini. Let's

18:43

download that one. Hold that thought.

18:45

All right. So, we got four or Pi 4 on

18:48

here. So, as you can see, it finished

18:49

downloading. So, what we can do is go

18:52

back to our chats or we can actually

18:54

click down here and load our model and

18:56

this will unload or eject the other

18:59

model and you'll see it start loading up

19:01

and you have the option to switch back

19:03

and forth if you want to. But what I'mma

19:05

do is not use the chat. What I'm going

19:07

to do is show you guys how to use it in

19:10

the terminal. Let's say you want to

19:11

start scripting. You can leave this

19:13

thing running in the background. That's

19:14

why it allows it to run in the

19:16

background like this. And then we're

19:18

going to actually I need to open it back

19:20

up. What we're going to do is start the

19:22

actual server. I want to show you guys

19:23

how to use API server. And actually, you

19:25

can do it from here. I just didn't think

19:27

about it. But yeah, start server on port

19:29

1 2 3 4. That's fine. But I wanted to at

19:32

least show you guys in here. So under

19:33

developer, we can start our server. And

19:36

boom, that'll start our service and

19:38

it'll share out all of our models. So

19:40

[music] let's go on, close that, and we

19:43

can open up our terminal. Let's get a

19:45

new window popping and move her to the

19:47

middle. And let's go on and zoom in a

19:50

little bit for you guys so you guys can

19:52

see. I already had a JSON written out.

19:53

What we're going to do is use the curl

19:55

command and we're going to curl against

19:57

our local host. And we're going to

19:59

specify our model. So, we're going to

20:01

use the fi and I've had this in scripts

20:04

for other stuff. So, what is it? Fi dash

20:07

let me look over there 4 dash mini and

20:12

reasoning is the actual model. And

20:15

basically what we're doing is doing a

20:16

curl of the local host. The port is 1 2

20:19

3 4 and then we're going to chat with

20:21

it. We're going to do a completion and

20:23

then context type is going to reply in

20:26

JSON format and then the model. We

20:29

specifying the model and then the

20:30

message, the role, user, and then the

20:33

content. Just you can put whatever you

20:35

want in here. You can ask the question,

20:37

whatever. You just have to put it in a

20:38

format. And so, let's go down, press

20:41

enter. And this may take a while. Ah, we

20:43

run into an issue right here. It says

20:45

failed to load three reason. Let's see.

20:48

Error. Model loading was stopped due to

20:50

insufficient system resources. So, let's

20:53

open this up right fast. Let's go back

20:55

up in here and let's just make sure.

20:57

Let's see. Do we have the model road?

20:59

Yeah, it's loaded. Let's go back in

21:00

here. It's loaded. Let's eject it from

21:02

the chat cuz it shouldn't matter. You

21:04

don't need it loaded. It'll find it. We

21:07

don't need a chat loaded. It'll find the

21:09

model that we want and it will load it

21:13

with the actual command that we run. So,

21:16

let's press enter. Boom. And you can

21:18

look over here on Yeah, this is

21:20

something I probably should have brought

21:21

up so you guys can see, but as you can

21:23

see, it's running a prompt. And this is

21:24

the information from the it's showing

21:26

you basically everything that it's doing

21:28

that the model is doing. It's trying to

21:30

go through and basically run the model.

21:32

So running chat completion on

21:34

conversation with one message. So

21:36

sampling parameters total prompt token

21:39

19 and prompt processing progress. So

21:42

zero progress 100 prompt progress 100.

21:46

So we should see some results over here

21:48

in a second. All right. So it finished.

21:50

Scroll back up just so you guys can see

21:52

right fast, but it used that model and

21:54

this is the results and it returned it

21:57

in JSON format. You can use this in your

21:59

code. So, as you can see, it says,

22:01

"Okay, let's see what the user needs

22:03

here." The original message says that

22:05

I'm pi a math expert for Microsoft.

22:09

[music] And then there's a greeting, but

22:11

the problem statement isn't provided.

22:13

Maybe there was a technical issue. But

22:16

anyway, basically he comes at the end of

22:18

it, he responds or it responds, "Hello,

22:20

it seems like you're ready to ask a

22:22

question, but I didn't receive the

22:24

specific problem you need help with.

22:26

Could you please share the details of

22:28

the problem?" So, this thing is ready to

22:29

work. You know what I'm saying? So, you

22:31

can add this to your code. You know what

22:33

I'm saying? Using the curl command to

22:34

gather information that you need and

22:37

then put it into your code, whatever

22:39

you're trying to do. So, you can call

22:40

this thing. And this is the exact same

22:42

thing I do with Olama. I actually called

22:44

the API. I basically send it information

22:47

that I need and it sends me back the

22:50

results locally. All this stuff is

22:52

locally and [music] organize it or puts

22:55

it in a format that I need it and then I

22:57

use it how I need to use it. So that's a

22:59

pretty cool feature right there. Now let

23:00

me talk about a few tips if you're

23:02

planning to use LM Studio regularly.

23:04

First you need to start with a small

23:06

model. I recommend you start with Fi or

23:08

Pi Gimma or one of those 3 billion to 7

23:12

billion llama models. They're great for

23:15

learning and also watch your RAM. You

23:17

know what I'm saying? Models load into

23:19

RAM. And that's why I ran into that

23:21

issue where it said we ran out of RAM.

23:23

Especially with higher quantization,

23:26

it means higher RAM usage. Also try GPU

23:30

acceleration. If you have a Nvidia GPU,

23:32

even if it's one of those mobile Nvidia

23:35

GPU, like in a laptop, you can still use

23:38

that GPU. It's still better than

23:40

nothing. [music] You know what I'm

23:41

saying? LM Studios can take advantage of

23:43

it. And also, if you're planning on

23:45

using, I don't know, [music] like a if

23:48

you eventually use like LM Studio plus

23:50

OAMA setups, then a wired LAN gives you

23:53

the smoothest experience. And also,

23:55

[music] LM Studios can take up a lot of

23:57

disc space, especially once you start

23:59

downloading a bunch of models. So, just

24:01

be weary of that. If you download

24:03

multiple models, it's going to take up a

24:05

lot of space. Cuz as you can see, both

24:06

of those two models that I downloaded,

24:08

they were around 3 gigs a piece. [music]

24:10

So, 2.5 for one and 3 gigs for the other

24:13

one. So that right there is close to 6

24:15

gigs. Just showing you [music] that's

24:16

two models taking up a good amount of

24:18

space. And then once you start getting

24:19

bigger models, they like you can get

24:21

some that's 7 gigs, you can get some 14

24:23

gigs, which I have some on my llama

24:25

server that are like 14 gigs. So it all

24:28

depends on what you need it for and all

24:30

that good stuff. All right, y'all. So

24:31

that's LM Studio running on Ubuntu 24.04

24:35

installed from the app [music] image and

24:37

loaded with a model and ready to use

24:40

with your local API. It's a clean

24:42

interface. It's easy to learn and it's a

24:45

great stepping stone into working with

24:47

local AI [music] models, especially for

24:49

folks who don't want everything going

24:51

through the cloud. And again, LM Studio

24:54

can work alongside Olama if you want to

24:57

combine them. But that's something we'll

24:59

save for another video. If you found

25:01

this helpful, go ahead and hit that like

25:03

button, [music] subscribe, drop a

25:05

comment telling me what models you're

25:07

using or what tools you want me to cover

25:10

next. [music] Thanks for watching and as

25:11

always, keep learning, keep building,

25:14

and of course, keep it techy. Yo, what's

25:16

up y'all? Listen, if you've been sitting

25:18

there thinking about making a move, let

25:20

me tell you, tech is where it's at. I

25:22

don't care where you coming from.

25:23

[music] Whether you've got a degree, a

25:25

GED, or just pure hustle, there's room

25:28

for you in this game. You see, [music]

25:30

tech is more than just keyboards and

25:32

code. It's solving problems, creating

25:34

opportunities, and building the future.

25:36

You already have what it takes because

25:38

tech doesn't care where you start. It

25:40

cares where you're willing to go. You

25:42

could teach yourself Linux, learn

25:44

Python, break into cyber security, or

25:46

even launch your own app. And the

25:48

resources are out here for free. And

25:50

yes, you heard me, free. Now, [music]

25:52

yeah, it's going to take effort. You'll

25:54

have to grind, but think about this. The

25:56

time is going to pass anyway. So, why

25:58

not invest it in a skill that'll change

26:00

your life? I mean, tech doesn't just pay

26:02

the bills. It opens doors to freedom,

26:04

[music] stability, and generational

26:07

wealth. So, stop doubting yourself,

26:09

store small, stay consistent, and keep

26:11

building. Because this isn't just a

26:13

career, it's a movement. [music] And

26:14

guess what? You belong here. So, let's

26:17

get it because the future is yours to

26:19

build. Keep it tight.

Video Information

YouTube ID: 3vdesqPDYEo
Added: Feb 21, 2026
Last Updated: 2 weeks ago