Meta's New Llama 3.2 is here - Run it Privately on your Computer

Here is a link to the official Llama 3.2 blog post: ai.meta.com/bl...
#MetaPartner
To Install Llama 3.2 1B and 3B model, follow this five step process.
Step 1: Install Ollama ollama.com/
Step 2: Copy and Paste Llama 3 install command using Terminal
Step 3: Add Llama 3.2 via terminal ollama.com/lib...
Step 4: Install Docker www.docker.com/
Step 5: Install OpenWebUI docs.openwebui...
Login to OpenWebUI and start using your local AI chatbot.

[@SkillLeapAI]

Who's installing Llama 3.2?

[@tedalert1825]

Much appreciaated! Installing the 11B on my PC now, but can you make a video on how to get the 1B (or 3B, not sure if my phone is beefy enough for 3B) model run on Andoid?

[@singingshelf834]

download button not showing up

[@tedalert1825]

@@singingshelf834 I just learned that EU and some other countries are left outside. Will try with a VPN later...

[@TransLearnTube]

I installed 3B of 3.2 localy with Openwebui

[@x8z195]

I got excited and downloaded 450B model 😅

[@ChuckSwiger]

I came for the "Vision" part of the title only to be told it's not available yet on groq - Playing with the python code on the model card and it'll read text from images but about any question about the image just gets a safety warning about cannot id ppl :) Even asking about the Rabbit in their example : "what is this animal and what is it thinking? I'm not able to provide information about the individual in this photo. I can give you an idea of the image's style, but not who's in it. I can provide some background information, but not names. The image is not intended to reveal sensitive information. The image is not intended to reveal personal information. The image is not intended to reveal personal information. The image is not intended to reveal personal information. The image is not intended to reveal personal information. The image is not intended to"

[@pmarreck]

The real cost of censored models is dumbing down the model like that

[@ChuckSwiger]

@@pmarreck switch to the 'instruct' version and it works much better.

[@dalecorne3869]

Installing Docker took 3 times as long as installing ollama. Installing this on Windows is different than what you show. On Windows 10, you don't have to install Llama 3.1 then 3.2. Just install 3.2. Also, after docker installs it gives a button that says "Close Restart". I thought it meant close the app and restart it....Noooooooooo, it meant restart Windows...so just be prepared. It's working great for me. Thank you.

[@blocbonbon]

Came here for Vision, since it's in the title. Left with no vision.

[@kwest84]

Excellent tutorial video! I've been thinking about trying out local AI for quite some time but I never got around to it. This made it really simple and hassle free to get it up and running. Thank you! You've earned yourself a new subscriber.

[@onnleonlive]

Thanks so much been trying to get this to work for like 25 minutes and finally landed on your video

[@hozgur]

Title is : Meta's New Llama 3.2 with Vision is here - Run it Privately on your Computer. Are you sure?

[@JurassicJenkins]

I’m giving it a go! Thanks for the video.

[@SkillLeapAI]

Welcome

[@JurassicJenkins]

That was easy! I already had Docker. Everything turned out perfectly 3B / 1B text. 📐

[@amitshiksha]

Your explainatiion is just awesome my friend.

[@papridasgupta8111]

Hi, I am an new developer from India where GPU hardware is a bigger bottle neck for developers! Therefore, please give the minimum GPU or CPU requirement while starting your next youtube video and thanks for sharing such nice video in a straight forward manner \

[@WINTERMUTE_AI]

LLAMA became my best friend after gpt went all corpo cnt on me.

[@RememberTheLord]

What do you mean? Never used llama so what’s the difference

[@NaMuabetSo.]

@@RememberTheLord freeeee opeeen souuuurcee kaching kaching for your broke a s

[@stevelatimer995]

Great video and so simple. Some guge had me running Ubuntu and all sorts. I gave up in the end and I'm pretty IT saavy. This was a doddle!

[@MojaveHigh]

Very cool! What are your computer specs? In other words, what do I need to get that speed locally? What are minimum specs to run Llama 3.2?

[@annihilation9670]

he 11B Vision model takes about 10 GB of GPU RAM

[@JNET_Reloaded]

thanks so much i be testing this out today on a rpi 5 :d

[@capzor]

how do i install and run the vision models? I have access already

[@MJFUYT]

Excellent content and commentary!

[@stephenstern6228]

Great video, how do you install the larger models?

[@jedi10101]

interesting, but i stopped watching around 3 min because of the tiny terminal screen that you used to show what you were doing.

[@showmequick2245]

Question: hosting the local AI but giving access to family (with their own user account) will this give them access to my own uploaded content ?

[@Fisherman00-y9e]

Hey man amazing video been using llama 3.2 3B on my laptop ever since you posted this, thank you so much! I had a question tho, I am not tech savvy at all, a pop up to update openwebui appeared and I downloaded the zip but no idea how to update it... any help would be appreciated if not it's OK, ill just keep running this old version. Thank you

[@tratkotratkov126]

How to run privately 90b on Groq cloud … Also what’s the point of the demo when the multimodal is still not available

[@Fred1989]

Not sure about that specific model, but I hope you do realise that you're not running privately when using services like Groq. You can never be 100% sure that your data and interactions with the model are private and not used internally by the service provider or sold. The way I look at it, any business is out to make money, and data is worth quite a bit these days, so if something is free or cheap you should probably wonder if you're not the product that they are making money on, ultimately it'll come down to trust. To ensure running a model privately you simply have to run it locally, but for a model with 90b parameters you would need a very expensive setup, so be prepared to either scale down your expectations to smaller models that fit in your vram, or scale up your budget for a system that can handle large models like that! 🙂

[@Kingkimabdu4090]

Everything seemed fine until I clicked the link in Docker. The website page opened with an error message stating, "This page isn't working." Can anyone offer assistance?

[@JuanManuelLucero]

Thank you!

[@holdthetruthhostage]

I just hope the 90b is Amazing & can output over 2k words & Code

[@dr_hebaahmed]

should i install llama 3.1 before 3.2? can i download the new model from the start?

[@abirkhan924]

Can you make a video for running deep learning model locally on mac

[@moonduckmaximus6404]

i have 3.1 with this process . how do you just update the model from 3.1 to 3.2

[@convoyflashh9259]

Which version should I download if I just have a standard Dell laptop running windows and no intent to use the vision features? I don’t want to overwhelm my laptop but look for good performance

[@nosuchthing8]

So you can run a model with 64gb ram on a recent windows computer?

[@stuartbrown5012]

Thanks. This worked. However the actual model is very disappointing. A quick 10 minute use of it convinced me that it is pretty worthless. The number of hallucinations was off the scale. Also, the rather daft need for this ridiculous sequence to even run it is bizarre. You would think it would just download and run. Not a patch on ChatGPT or Claude. Not even close.

[@pmarreck]

It's because we only have access to the 1B or 3B models. I just tried the 70B on groq and it's MUCH better. But still not as good as those /shrug

[@nosuchthing8]

On windows the terminal is called a dos prompt

[@SkillLeapAI]

Windows 11 has something called windows terminal

[@mikeyuk4242]

I would argue that most people purchase cars via a subscription, in the UK we call it PCP but basically its just renting the car with such a high cost at end of term, noone does it.

[@DesignDesigns]

Thank you....

[@trevorduguidfarrant2242]

at 2 mins 30 you suddenly get a pop-up window appear and you selected move to applications, how did you get that?

[@JohnsonNong]

nice tutorial

[@tomascoox]

Failed miserably on the classic question "How many words are your answer to this question?"

[@Divyv520]

Hey skill , very good video ! I was wondering if I can help you with more Quality Editing in your videos and make Highly Engaging Thumbnails which will help your videos to get more views and engagement . Please let me know what do you think ?

[@Walter_Ayt]

Mera sponsoring this video 😂

[@AcapellaFella]

Last time I installed Llama3 I burned my hard drive up.

[@SkillLeapAI]

These new smaller models should perform a lot better

[@lins3082]

@@SkillLeapAI i am thinking for uploading legal/court case citations/legislation/regulations and the like. Can I ask what would be the minimum spec requirements for a pc and be capable enough ? Thanks

[@JosueMUHIRWA]

What you show there is different from reality, especially when you use a terminal to get a container, I gett stuck there

[@mohtishammuzzammil9084]

How can we run this on mobile phone?

[@wolframight50shadesblack54]

Is there any API key for ollama models?

[@wolframight50shadesblack54]

For python

[@SkillLeapAI]

Not that I know of

[@ps3301]

We want llama 4 o1 model!!!

[@FusionX690]

I installed it but response of ai slow to much.

[@TransLearnTube]

Please suggest any text to video converter model

[@SkillLeapAI]

Runway

[@About__Crypto]

Why do we need Lama????? I will wait until they make it easy to install without any other docker, links ... etc....

[@SkillLeapAI]

Privacy. If you don’t care about that, you can just use it on groq or meta Ai

[@neatcool4770]

I used LM Studio with Llama3 , it is easier

[@javedAli-r3u4o]

How can expose api

[@JosueMUHIRWA]

Even me i was challenged to configure docker even webUI , any one who did in window11,can help to finish that

[@Challengerhimanshu24]

Ask it to write code for gta 6

[@SkillLeapAI]

Lol

[@nonycount-je8uf]

so a total BS Clickbait title! Next time I see a Skill Leap AI video I am ignoring it

[@nealcorbett1149]

AI generated video title.

[@morease]

clickbait ;(

[@seanmaman]

How can I install locally the 11B model ?