In short
OpenAI’s release of two open-weight AI reasoning models, available to download on Hugging Face, and what it means for individuals and businesses (including self-hosting and “deep research” style workflows).
Guests/backgrounds
The episode is hosted by Jamie and Jaeden (AI Hustle School). Jaeden runs a startup, AIbox.ai, offering access to ~40 AI models for about $20/month.
Key claims
OpenAI’s models are “state of the art” among open models; you can run them without OpenAI API access or paying API credits (subject to licensing). Larger model (GPT-OSS 120B) can run on a single NVIDIA GPU; smaller (20B) can run on a consumer laptop with ~16GB RAM. Open-sourcing may be goodwill ahead of a rumored GPT-5.
Notable examples
Potential uses include analysts/finance reporting and lawyers running locally to avoid sending confidential client info to the cloud.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Chapters
Tap a time to open that second in VOInside the AI Hustle School Community
0:45 to 2:30
Discussion about the exclusive content and benefits of their school community.
“There's a really great kind of community where people are talking about it.”
OpenAI's New Open-Sourced Models
2:30 to 4:48
Exploration of the significance and implications of OpenAI's new models.
“basically device, computer, or server, or whatever.”
Use Cases for Open Source AI
4:48 to 7:19
Potential applications of OpenAI's open-source model in various fields.
“And they have a lighter weight model, which is a 20B, 20 billion parameter model.”
The Competitive Landscape for OpenAI
7:19 to 9:19
Analysis of the competitive dynamics surrounding OpenAI's open sourcing move.
“Sam Altman in January of this year said he believed that OpenAI had been, quote, on the wrong side of history when it came to basically releasing some of this open source stuff.”
Transcript
Automatic transcript. May contain errors.0:00Today we have some big news out of OpenAI. They just launched two new open-sourced AI reasoning models, which is pretty big news. We're going to talk today about some of the specifics of that, sorry, and get deeper into what that means for you. But before you do, Jaeden, why don't you tell them about our school community? Yeah, every single week Jamie and I actually record a video over on our school community. We post it there, we don't post it anywhere else. And we're basically breaking down how we use AI tools to grow and scale our businesses. So they're like literal demos. We do walkthroughs.
0:33We show numbers of how much money we make from different things. Basically, it's everything we can't post publicly. We put it inside of this group. We have over 300 members of the group. And so everyone basically gives feedback. There's a really great kind of community where people are talking about it. This week, we recorded a whole video on tips and tricks that I've done with deep research, how I'm leveraging OpenAI's deep research to accomplish a bunch of different things where I think you should and shouldn't use it. I showed demos. I had an interview yesterday with the Wall Street Journal on deep research.
1:05And so I pretty much took all of my like best advice that I shared with them and condensed it into a video that I'm sharing before that article ever gets published over here. So if you want the inside scoop and all that kind of all the inside tea, go check it out. The link is in the description for the AI Hustle School community. But yeah, let's get back over to what's going on with OpenAI. This is a phenomenal announcement. I think people have been harassing them for a long time. I mean, endlessly about not being an open source company anymore because everything's closed source and Elon Musk suing them and saying that they should be called closed AI instead of open AI.
1:41Like there's just unlimited drama. So this is really exciting for open AI to finally be dropping their first in ages. I mean, years. This is their first what they're calling open weight AI reasoning model. so basically um it's uploaded to hugging face you can go actually download it um on hugging face and it's they're saying it's state of the art when you state of the art when you're measuring it across a bunch of different benchmarks if you compare it to other open models obviously the newest models out of a google and grok and open ai and anthropic are like better than it but it's the best open source model the best model that you don't need their permission you don't need their API.
2:20You don't have to pay them to run this. If you have good enough hardware, powerful enough hardware, you download this model, they give you the weights, and you can just run this on your own basically device, computer, or server, or whatever. And so companies can actually use this, although there's always kind of licensing requirements and rules about this. But really, really exciting that they're finally releasing an open source model again, which yeah, it has a lot of really cool implications to come along with that. Yeah. So I'm trying to think of the significance of this for me as a normal person, or even thinking about how a company could use the open source model of specifically reasoning.
3:01I guess, what are your thoughts on that? Because my mind is like, why would an app need to do deep research unless you're working in some kind of, if it's a research app or if it's like a study app? What are some of the use cases that you see for this? I mean, the deep research element, personally, I actually think this is interesting because they haven't really made API access for deep research available to a lot of most of their business consumers. So basically, if you have the API to open AI, you just get access to their models. You don't get access to all of their tools. So having the deep research on this is actually kind of cool, in my opinion.
3:36For example, my own startup, AIbox.ai, where I have 40 of the top AI models on there, and it's like 20 bucks a month you can use them all it's like cool but sometimes there's these tools like deep research that we can't offer stock from open ai we try to do our own software layers on top to to add like some features you'd see other places we also have tons of other features that they don't have that's kind of why we built the platform and you get access to everything for one price but i'm excited when they have something like this come out because we have a vendor we use for open source models that will add this to it and then all of a sudden we weren't getting access to deep research before.
4:11And once this comes out, we'll get like open AI is built in deep research, which is kind of cool. So use cases, I think there's a ton when you want to do research on anything. I think data analysts are going to find this really interesting, all sorts of analysts and people doing reporting for big organizations, people in finance. I think there's a lot of use cases that are interesting. And to that point, like anyone can run this. They have two sizes that the model comes in. So the biggest is going to be, is their GPT-OSS 120B. And this can run on a single NVIDIA GPU. So basically, if you have some good hardware, you can run it on a single GPU.
4:49And they have a lighter weight model, which is a 20B, 20 billion parameter model. And this can run on a consumer laptop. So assuming you have 16 gigs of memory on your laptop, you can actually run the lightweight model on your laptop. That is so cool to me. So really excited. that a lot of people are going to be able to just start running this stock, basically. That's cool. So I guess I'm picturing like even a lawyer could have it do research, deep research on past cases and kind of, you know, pull information and compile like a list. Is that kind of what you're what you're saying? Or an analyst could pull up, you know, the companies, how much they spend for their break room coffee each year, you know, and stuff like that.
5:30Is that what you're you're kind of getting at? Yeah. And if I'm being 100 % honest, I don't think I'm getting we're gonna see and so you're capable of it and I think it's exciting and interesting I don't think we're gonna see a lot of lawyers using it just for like fun although perhaps people that are doing like very highly confidential things a Sam Allman recently said like put confidential information into chat GPT because people are like using it like a therapist and he's like don't talk to like a therapist because it doesn't actually have any sort of legal protection if the police had a warrant they could come to open AI and say give me all of their chat logs and they could get all of your chat logs from open AI.
6:07And so he's like, you know, if you had some sort of personal information, you didn't want us to share with somebody, like don't put it into chat GPT. And a lot of people kind of freaked out about that. This would basically solve that problem. If you're running one of these open source models locally on your computer, whatever you type in there is not going up to the cloud. So if you were a lawyer and you had sensitive confidential information about your client that you were like putting in there and personal information to try to help you build a case, it would be much more beneficial to use something like this locally running than to have something that's sending information over to OpenAI.
6:38So that's like one potential use case. But I think the most popular use case basically is going to be for businesses that are basically be able to self host this without paying for the API and API credits to OpenAI. So you probably could save money if you had your own hardware to run this basically. Okay, cool. Yeah, I know that that makes a lot of sense. Thanks for clarifying. Yeah, it's kind of interesting. Literally, this is a big deal because this is the first time in five years, the last time OpenAI released a model was GPT-2. So this was like way before chat GPT even launched. They had something called GPT-2.
7:12They had kind of an open source model more than five years ago when they launched this. So this is the next time that they've done this. So it is pretty exciting. Sam Altman in January of this year said he believed that OpenAI had been, quote, on the wrong side of history when it came to basically releasing some of this open source stuff. And Elon Musk criticized them endlessly. A lot of people did. They have a lot of pressure from other companies in China. DeepSeek, Alibaba's Quan, Moonshot AI, all the Chinese labs are releasing their models open source as well. And so it basically makes OpenAI look pretty bad.
7:46And so I think this is kind of why they pulled this out. Now, that being said, though, open sourcing this model means that all of their competitors, theoretically, a lot of the Chinese companies like DeepSeek can grab it and use it to train their own models. And so there is this whole competitive reason why they've kind of tried to stay away from it. But I think, basically, if I'm giving it 100 % honest, I think they're launching it now to build Goodwill. And they probably are doing this on the eve of a new updated model, KoffKoff GPT-5, that is going to blow everything out of the water. So they're like, eh, whatever.
8:19For all of you peasants, here's your open source model. It's the best right now. So we'll get all the praise for releasing it, have fun with it and then in two weeks they're going to release something insane that like basically makes it useless and that's my opinion of what's going to happen no that's i mean that's a great that's a great thought because i had i had heard rumors that gpt5 was to be expected um you know this summer so that would then timing wise so that would make sense and you're probably spot on with that honestly yeah so it's going to be interesting to see how this works how does it perform pretty decently compared to a lot of other open source models um i think uh it's going to be it's going to be interesting to to see where it goes but again it's good now and i don't think it'll be good in one like there is no way in my mind with all of the rumors that in one one to two months gpt5 isn't out and possibly it's going to be out like in one to two weeks so with all those rumors uh yeah take out the caveat that there's probably something much much better on the horizon.
9:20So, you know, can't clap too much for open AI. Hey, thank you so much for tuning into the podcast today. If you enjoyed the episode, make sure to leave a rating review wherever you get your podcasts. And if you want to learn how to grow and scale your business or your career using AI tools, make sure to go check out the AI Hustle School community. There's a link in the description. And yeah, we'll catch you in the next episode. Hope you all have a fantastic rest of your day.
From the publisher
In this conversation, Jamie and Jaeden discuss the recent launch of OpenAI's new open-sourced AI reasoning models. They explore the significance of this development, its implications for businesses, and potential use cases for the models. The discussion also touches on the competitive landscape of AI, the future of OpenAI, and the anticipated release of GPT-5.
AI Hustle YouTube Channel: https://www.youtube.com/@AI-Hustle-Podcast
Our Skool Community: https://www.skool.com/aihustle/about
- Try AI Box: https://aibox.ai
- YouTube Video: https://youtu.be/CQvbPNGN7Uc
Chapters
00:00 OpenAI's New Open-Sourced AI Models
02:49 Implications of Open-Sourced AI for Businesses
05:32 Use Cases for Open-Sourced AI Models
08:32 Future of OpenAI and Upcoming Models
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.




