OpenAI Open Source Models Are Here - Run GPT-OSS Locally on your computer

I tried it for like general use and it was decent, but it started making up stuff for my niche use case, which all other local LLMs fail at as well

Could you please explain how to make ollama or LM Studio working with local files by creating, editing, deleting, etc for coding purpose?

you need to go to setting on ollama and enbaled cloud to downlaod it now

After downloading the modal I still need to pay?

thanks!

where to find what would be the perfect system config to run the LLM locally ?
I look to deploy a VM and start exploring there

Great vid! Does LM Studio have the web search function by any chance?

Can I run 20b with 24 gb ram?

if i only have gpt team can I use it with no extra cost? or will need to pay for api?
I want agent to be in cloud and just to have control over my pc

What device are you running gpt oss on? and what are its specs?

Thank You. Downloading Lama ans will try chat gpt. Question, can I use multiple AI, like deepsek + chatgpt on lama?

Had a reallly bad experience with Ollama. It did not load the model in the GPU properly for me.

Where is the value of reading a web page?

So olla a does not allow to attach files but lmstudio does

Any open source models you recommend to analyze media files offline?

How do i get the dark theme for it on windows? i dont want my eyes to burn out before the ai apocalypse

Why does no one asking the obvious question!!! Does running it on my laptop lets me BYPASS the woke censorship and do i get an answer to ALL questions? I mean isn't this the whole purpose? If not then i could just use the official gpt with all the features.

0:33 min Can a laptop with 64GB of RAM run the 120B parameter GPT-OSS model locally?

Apple macbooks are not made for LLMs

That was impressive man 👏 now tell me lets say we have downloaded the open model of gpt in our computer now whats the benefits apart from privacy... Since privacy is not my concern

Can it create images?

Looking forward to the video showing this running on Windows

You can simply search and download HF models directly from the LM Studio UI, you don't need to run the terminal

can you make the video of differnce between the 120b and 20b .
and make a video of showing 120b soon

Thanks for the guide, but one question, should we use these open models for our company, since we have internal and sensitive data? Any data leaking potential?

Why doesn't Ollama work without Wi-Fi?

Cheers for sharing, Raj! 🏆

U didn't show whats it capable of, just throwing a video to be first.

🎯 Which GPU do you have and how much vram?

Without internet connection it doesn't work even after installing

Thanks for this. Now running well on an M4 Mac Mini Pro using LM Studio. It does need some pretty detailed prompting to get worthwhile outputs, but worth checking out.

Honest question - what practical use case is there to run these LLM's locally? Are there examples available of how people are using these?

Thanks, installed it in a jiffy with your help.

Thanks!

Thanks for sharing! Always keen to try local llms!

Hey Skill Leap ai ,awesome video—really enjoyed it! I’m a professional video editor and thumbnail designer, and I help creators like you boost views and engagement with high-quality edits and scroll-stopping thumbnails. I offer fast turnaround times, great pricing, and a custom approach for every creator. If you’re open to it, I’d love to show you how I can take your content to the next level. Interested in a quick preview?”

Quite impressed. the 20b is one of the few LLMs that knows who the 3rd president of Finland is. Most models get it wrong or make up names, until you ask them to list them, the ask it who's third on the list.

Could be instales on a VPS?

GPT OSS being free and able to run on regular hardware is huge I’m curious to test it with Genum AI managing prompt flows to see if it can really match the smaller GPT 4 models in practice

How to get this on Linux?

Looks like that second option you showed briefly has the ability to upload files/pictures to the model, can anyone confirm? If so that would be the preferred method I feel.

This video could be 2 minutes long

is it good for devlopment ?

not so open.. it is riddled with censorship.. the poor thing is afraid to think..

locally with nvidia rtx 4060 and i714650 hx, responses are coming slow.. is anything required to be done while using ollama app ?
also, can it be used like agent mode in vscode the way github copilot is used ?

You can run it on GCP

Dude.. this is false, when you ask "gpt-oss:20b" if it's the open source model it literally says it's chatGPT 4