OpenAI Releases World's Best AI for FREE (GPT-4o)

So it happened again. OpenAI just smashed the ceiling on what we thought was possible with its brand new GPT4O model. This video is going to give you an overview of all the things that came out, including the extremely deep blog post. There's a lot hidden in there that was not covered in the live event. Plus I'm going to talk about some of the implications that might be not so obvious because I just spent the past few hours Talking to other YouTubers, YouTube community members, AI Advantage community members, all about this brand new model that is the new gold standard when it comes to AI as a consumer. And trust me if I tell you there's a lot here. This is one of those moments where all these exaggerated YouTube titles are more than warranted because OpenAI just changed everything again. And here's what you really need to know. Okay, so first things first, they just kicked in the door one day before the Google event happening tomorrow with their YouTube livestream titled Introducing GPT4O. Now the O stands for omnimodal. This is their word for multimodal. So now we have GPT4O. What's new, you might ask? Well, first things first, it performs better on all the benchmarks than any other model, including their previous GPT4 model. But that just scratches the surface because it's also much faster. It's two times faster than the previous version in English and it also has improved capabilities in 50 other languages as outlined here. Now here's the interesting thing that they did not talk about these improvements stack. So because it's 2 times faster in English and around 3 times faster in Hindi, in practice, it's actually 6 times faster in Hindi because these stack. And look at all these improvements. It was decent in other languages, now it's great. So those are some of the objective stats. But as you might know, a lot of the power of ChatGPT lies within the subjective experience. How good is it for you for your use cases, for your everyday usage of this application? Well, GPT4O is now their omnimodal model, which means it has better performance across text which we had, audio which we also had, and images which we had. But they are adding capabilities for it to take in real time video from your phone too. So they will be shipping brand new applications not available today. We'll talk about that in a second. But you'll have an iPhone app and a desktop app that can either use your camera to scan the real world and interact with it in real time with the new Voice AI assistant which we'll Also, talk about in a second here, they completely revamped the voice capabilities and with that voice assistant, it's going to be able to see your desktop, it's going to be able to see the world around you with the phone on your camera and and assist you that way. And this was simply not possible yet as there were massive limitations on the vision capabilities. You could upload one picture, then it took a few seconds and gave you a half baked answer. From the various demos they have shown now and they were two in the live streams and there's more than a dozen demos here on their website. It's actually a massive improvement to what we had up until now. So before getting into all the details, just check out this demo where Our father uses ChatGPT on an iPad next to a math problem. And the app essentially guides the student through how to solve this. And it's a tutor.

Can you first identify which sides of the triangle are the opposite adjacent and hypotenuse relative to angle alpha? All right, so I'm pretty sure this is the angle alpha right here, right?

Correct.

Correct. Now, looking at the triangle, which side do you think is the hypotenuse? I think it might be this one, but I really am not sure this side. A.C. you're close.

So as you can already see, there's a lot of differences to the current version. First of all, you can actually interrupt the AI assistant. Secondly, and maybe even more importantly, it is snappy. It answers like a human being. There's no answering and waiting for four to seven seconds for it to come back with a response. Plus, it's taking in the iPad screen in real time, something that just simply wasn't possible up until now. This is a feature that is going to ship over the course of the next few weeks to plus users. But. But this entire GPT4O model might be an even more exciting announcement to free users because every single user will get GPT 4.0 for free. That includes voice input, GPT 4 level intelligence GPTs, all these premium features that were paywalled behind a $20 a month paywall up until now. Everybody will have GPT 4 in their pocket now. And finally you will be able to share GPTs with other users. GPTs that are powered by this more capable model. This is huge. If I wanted to show family members some things that I've built or done, or maybe make their life easier with some of the skills that I have taught on this channel over the last year, I couldn't. They needed a premium account and that is a barrier that you're just not going to cross with your Grandma. And the $20 a month is just something that most people did not even consider in their everyday life. Now you can show them the power of these tools firsthand. But look, the base facts that I covered here just scratched the surface. There's more. As I mentioned, they uploaded the dozen videos with very various use cases. And you can actually go to this part of their blog post, which I'll link below, where you can look at all these different improvements. And some of these are super unexpected, like 3D object synthesis. Look at that. The model can do that now. Also, the model can create fonts, which means they also solve text. There's so much more. I will turn all this and everything that the community and the Internet brings up into a separate video on use cases of this brand new model. So what else is new here? Well, the fact that it has better vision sounds like something very subjective. But in a live demo during the event, I was actually very impressed with one particular part of this scene. Yes, it's a fantastic live coding assistant. We had versions of this before. If you really wanted something like this, you could make it happen. But this example where he uses his phone and he actually shows emotion and chatgpt4o correctly identifies what he's feeling. This one in particular really blew my mind.

It looks like you're feeling pretty happy and cheerful with a big smile and maybe even a touch of excitement.

And here's my point. All other models could pick up on basic emotions with they had vision capabilities, right? You could send a picture to ChatGPT and it would identify what emotions you have, but only at a very high level. If you look at the full spectrum of human emotions, and there's many versions of this wheel, I'll just be using this one. Usually it only managed to identify the first circle here. Okay. So it could tell you, hey, you look happy, you look sad. But it never managed to hone in on the details here. As you could hear here, it told them, hey, you look pretty happy with a hint of excitement. Now, as you can see, excitement is actually in the second wheel here. And up until now, these have been pretty much impossible to identify for most AI models. I'm going to be testing and comparing this extensively, but that will cover in another video. So subscribe to the channel to catch that video. But moving on, this is not the end. There is more here because it's not just smarter, it can talk to you in real time, can see what your smartphone is seeing. Or what you're doing on your desktop, but it also has better perception of emotions and all of that and more happens faster. But don't take my word for it, let's hop into the web interface that by the way also slightly changed as you can see here. Now it's more of a conversation with chat bubbles. This is old GPT4 answering in real time as I'm talking. This thing is generating. You can get a feeling for the speed yourself. Now I'm going to stop this generation and switch on over to GPT4.0, which by the way is rolling out as of now. I talked to the Advantage community and most people actually have it already. So if you're A plus subscriber, chances are you have it because I'm recording this from Europe and that means I'm usually at the end of the queue of these rollouts. So let's just do the same thing in GPT4O you can instantly see the difference. This was more than a 2x speed improvement. That's what they promised, but it's done generating. Look, I didn't even manage to finish my sentence here and as you can see there's new features here where you can switch the model and regenerate. I could regenerate in 3.5. Basically it allows me to mix and match models as I go, like so. And that brings me to an important topic which is now that everybody has free access to this, why should you stay subscribed to the plus plan? Well, this was one of my main questions and there's two really good reasons. First things first, you have five times higher rate limits than the free users, which some members in our community actually did the research and said it's about 80 messages per hour right now. So this allows real GPT4 usage where you're not going to run into the limits easily. You going to be able to fluently interface with it and just get the responses. This is even more relevant as the GPT4O model is way faster at generating. But the second reason I think this is the main one, all of the amazing vision features where it uses your screen, where it uses your phone camera as context on top of the prompts, those are only going to be rolled out to the plus users. These are not available today. They said they're coming over the next few weeks. That is all the info we have right now. In typical OpenAI fashion, this will probably mean a gradual rollout and and by the way, to all API users, The API on GPT4O will be 50% cheaper than what we're used to with GPT4 that is also available today in the Playground and Food API. You can simply go to this link and here you have the GPT4O option again, Lightning fast generation. And it can do something like writing a simple Snake game in Python, as most top models can do these days. But look at that. I mean, this took 10 seconds to generate. And once we have the voice control, I will be pretty much going through every single app and workflow that I do on an everyday basis with my GPT4O assistant open. And I'll be sharing my screen and trying different workflows and see if it can assist me or come up with new ways to improve my productivity with these tools. And I think this is really the key here. This really opens up the superpower of enhancing workflows that you already have, making the applications you have more powerful. As I said so many times, AI is the next evolution of software and this is the form it's going to be taking. You're going to have an assistant on your desktop, on your phone, that is going to be aware of your surroundings, of your work that's happening right now, and it's going to be helping you along the way with all of its knowledge and experience and the context that you provide to it in advance. But just like you're raising your awareness of this tool by watching this video, it's important to be aware of some of these use cases and the possibilities that this hides because these tools are deeper than they seem. So at this point I would be curious, what are you excited about doing with these new tools? You will have access to this within a few weeks. What is the first thing that you're going to try? What part of your life do you want to improve? Leave a comment below with that. And and matter of fact, I asked this exact question in the AI Advantage community and people started to share how it's going to affect their various professions, and then they brainstormed together on how they could take it a step further. Most days of this week we'll have events doing exactly that. Tomorrow I'll be covering the capabilities and use cases that I find. So if you're interested in discussions like that and really taking the next step on educating yourself about AI, you can check out what me and the team built under aiadvantage.community. but if you want to simply explore what this technology holds for you and stay informed, YouTube might be exactly the right place. Matter of fact, I actually have an extra recommendation which is another YouTube channel that we're doing with some other AI YouTubers, and we just held an 80 minute livestream of watching the event together and most importantly, sharing our first impressions. So if you want the first impressions in a live setting between multiple AI YouTubers, you can check that discussion out over here. And other than that, I'll get back to exploring what this can do. And once I discover more, I'll be reporting back on this channel. Oh yes, and for anybody asking what's up with GPT5, all they said is the next big thing is coming soon. Whatever that means. We shall see. But this is fantastic for now.