GPT 5 News - Everything We Know So Far

Where is GPT5? This is the infamous model that is likely going to be a large step up in terms of capabilities that many are expecting from OpenAI. Let's actually dive into actual facts and not any speculation about what the model is going to have and all the key details. So, one of the key details that we do know is that GPT5 does exist. It's not this speculative model name that we don't know whether or not we're getting it. 100% it does exist. someone who recently left OpenAI as we can see here actually said that they've worked on GBT5 and they helped train GBT5 and future models. So, it's quite likely that GBT5 and future models have already been trained which means that they're quite nearing the end of their development and are soon going to be deployed. So, the reason I've said this is because I want this video to be grounded in reality and not based on any speculative rumors. So, someone literally just updated their resume with this. Now, of course, remember we actually did get this tweet from Sam Alman where he basically gave us the road map for GPT4 and GPT5. Now, this road map here was actually quite early and he explains it in detail. So, this was the road map that was posted in February the 10th of 2025. He basically said, "We want to do a better job of sharing our intended road map and a better job simplifying our product offerings. We want AI to just work for you and we realize how complicated our model and product offerings have gotten. " What he's talking about here is the fact that currently the way how you use AI models is quite confusing. You've got GBT40 mini, GPT40, GPT4. 5, 01, 03, 03 mini, 03 high, 04 mini. Seven different models for many different use cases. And the average person is just getting decision fatigue. There's too many models. And this is actually tying into how a GP5 will work in the future. It's basically going to be a unified model. So he basically says, "We hate the model picker as much as you do, and want to return to Magic Unified Intelligence. We will ship GPT 4. 5, which they already have done as our last non-chain of thought model. " So remember now there's never going to be another solo model. That's the last non-chain of thought model. What they're going to do is basically unify the O series of models and the GPT series of models by creating systems that can use all the tools. Systems that know when to think for a long time or not and generally might be useful for a wide range of tasks. In both the chat GPT and the API, they will release GPT5 as a system that integrates a lot of their technology, including 03. They do say that they will no longer ship 03 as a standalone model, but remember this was earlier this year, and they've already shipped 03, but they did say that the free tail chat GP2 will get unlimited chat access GP25 at the standard intelligent setting subject to abuse thresholds. I highly doubt this will be the case once it's released because I've seen OpenAI models on release day. They never really have enough compute to house the model because there's so much hype and they always underestimate exactly how much compute is offered. So I would highly doubt that about GBT5. They state here that plus subscribers will be able to well this is Sam by the way will be able to run GBT5 at a higher level of intelligence and pro subscribers will be able to run GBT5 at an even higher level of intelligence and the model will incorporate voice canvas search deep research and more. Essentially, GBT5 is not just going to be your standard chatbot. It's going to be an AI that has access to several tools. Being able to create images, create videos, search, voice, deep research, basically do many different things for you, quite like your virtual AI assistant, that is truly capable. And in terms of how you're able to use the model, it's really going to depend on your level of subscription. If you have a standard subscription, your model probably won't think that long. But the more you pay open AAI, the longer it's going to think about your problems and thus the accuracy of your answer is going to increase. So now, just as we were talking about that, I think this stuff is starting to get my brain working. And if you enjoy that feeling, too, you'll probably want a fun way to sharpen your problem solving skills every day. And if that's true, you have to check out today's sponsor, brilliant. org. Now, Brilliant helps you get smarter daily with thousands of truly interactive lessons in math, science, programming, and data analysis, and of course, AI. Seriously, their catalog is huge. Now, what makes Brilliant so effective, and this is why I personally love it, is the hands-on approach. You're not just watching lectures. You're actively solving problems and playing with concepts. It's built on first principles, helping you understand why things work and not just memorizing. Now, this method is actually proven to be way more effective for building real understanding and critical thinking skills. Skills that apply everywhere. Seriously, stop mindlessly scrolling and build a powerful daily learning habit with. So, whether you want to finally grasp calculus, understand the logic behind AI, build your Python skills, or make sense of data visually, Brilliant has engaging courses designed by experts from places like MIT and Google to get you there. Now, you can try everything Brilliant offers for free for 30 days. Plus, you can get 20% off an annual premium subscription by visiting my link on screen. You can also scan the QR code on screen as well, and the link right there is in the description below. Don't forget to check it out. Now, what is interesting is that Samman actually spoke more about GBT5. He actually gave

us some more key details. He actually said, "Change of plans. We're going to release 03 and 04 mini after all. Remember, we already know this. " And then, of course, he said, "We're actually going to do GB5 in a few months. " Now, this was tweeted in April, so a few months could be all the way up until the end of the year. and he said there are a bunch of reasons for this but the most exciting one is that we are going to be able to make GBT much better than we originally thought. So whatever happened at OpenAI they may have had some breakthrough that allowed them to find a way to make GBT even better than they had. I'm not sure what breakthrough that is. We do know that them delaying it means that they're trying to incorporate those capabilities into the model. And they did state that unifying these models is a lot harder than they thought. So smoothly integrating everything is taking a little bit more time and they want to make sure that they have enough capacity to support when they expect that unprecedented demand. Remember how I literally just spoke about the fact that every single time a model gets released, you really don't even have the ability to use it because everyone and their grandma is trying to. Of course, right now, one of the key problems OpenAI has is they're still trying to get enough compute so they can serve the models. I do think further competition will, you know, force them to release models early or even do certain demos early, but of course this will factor into the time of delay. Now remember, we also got Kevin Well, the CPO of Opening Eye on an interview actually talking in detail about GPT5 and how they're working on it. Yeah. Well, we're very excited about GBT5. There are teams working on it as we speak and I think it's going to be a fantastic uh new model when we get there. But in general, this is the year that chat GPT goes from uh answering questions for you to doing things for you in the real world. Uh so, you know, by that I mean we have we launched a product earlier this year called operator that can browse the web like a human and do things for you. When I think about the time I spend on the web, I bet 30 or 40% of the time that I'm browsing the web, it's not because I want to be doing what I'm doing. it's because I have to get something done and there's some series of pages and you know forms I have to fill out and other things. I would love an AI to do that for me. I would absolutely love for an AI to fill out those forms for me so that I can concentrate on, you know, either getting more work done or spending more time with my family. And this is the year when AI begins to do more things like that for you. Uh and I'm very excited about that. It comes because of the advances that we're making with these reasoning models that are getting smarter and safer and it comes with you know integrations into the sites that you know and use every day so that our models can understand them and can take action on your behalf. Now he actually responded to a comment here where he spoke about the update and someone actually asked in GPT5 are GPT and Oer still separate models under the hood or are you making a model router? essentially just asking are you basically having one query and then having it routed to different models under the hood or are they going to be unified in some more substantive way. He actually says here that the models are going to be unified. That means it's not just going to be a model router that picks the best model to choose from. It's actually going to be completely integrated. Meaning it's going to be one entire AI system that's built differently in a way that allows it to reason about different problems and internally just select them. So this is going to be a lot different from just a model router. And that is something that truly intrigues me because I know exactly what a model router is, but I am going to wonder what kind of system they're building that initially can intuitively pick which parts of its brain essentially to use to kind of reason about the task. I think this might be akin to an actual AGI. And I know that does sound crazy, but when we think about what AGI is and we think about the human brain, we think about things like language or vision, speech, these are essentially going to be things that the GPT5 series has. If GPT5 is an AI model that can speak naturally, has the ability to see, can of course output audio, can digest videos, of course, can write text, of course, it's a digital thing/s system, but I think that may fit some kind of definition and of AGI if you guys understand what I mean. So them actually saying this is unified is a clear detail that GBT5 will be levels above. Now, of course, right here you can see he says, "Can you elaborate on this if possible? I interpreted your reply to mean that GPT5 will be an omnire model as opposed to a router between an omni model and a reasoning model, but I feel there are some conflicting signals and would appreciate more clarity. " He says, and this is Kevin Will, CPO of OpenAI. What you outlined is the plan. We may start with a little routing behind the scenes to hide some lingering complexity, but mostly around the edges. The plan is to get the core model to do quick responses, tools, and longer reasoning. So overall, it's going to be a unified model experience. Now, as for the release date, some people are speculating that with Google IO coming up, Google may once again raise the bar in terms of what is acceptable in AI. We already know that they've done that with Gemini 2. 5 Pro, and it's really caught OpenAI off guard. So many people are

thinking that since OpenAI usually shows up Google's events like the day before or the day after, they usually release an AI tool that gets everyone engaged in theirs and basically skips over Google that maybe in May we may actually get something from OpenAI. So that will be a key date to look out for. Now in terms of the parameters for the model, we actually did get this from a Samsung exec at Semicon Taiwan. And I'm not sure if this one is confirmed or not because this was actually in 2024, but we can see here that the parameters for GPT5 was actually 325 trillion parameters. And we can see it was trained on the B 100 series. Now, like I said, I'm not sure if this is true, but this is a really reputable source. This is the corporate president and the head of memory business at Samsung Electronics. So, I would argue that this is probably as legit of source you're going to get. But maybe this is, you know, changed and maybe this information wasn't meant to be out there cuz I really didn't see this anywhere else. Actually saw this on Jimmy Apple's Twitter. Of course, as most of you know, Jimmy Apples is essentially an infamous AI leaker. Now, if we're talking about the jump from GBT4 to GBT 5, Samman actually says that jump is going to be rather incredible. I think, you know, going from GPT3 to GPT4 really surprises the world. um GPT4 to GPT5 will be a similar leap and uh it's amazing capability to these models they're getting so smart that maybe by the time of GPT 5 or six maybe you'll say like that's enough intelligence it's already smarter than me at this point how do you make it how do you integrate easy to use can we bring voice mode and canvas and video and all of our features together and can we have just this one model that can do everything that knows when to search the web go to a research project. It knows when to write code. like switching the voice mode. So I think that integration will be the next big milestone after we gain the intelligence. Uh we want to bring GPT and O together. Mhm. So we have one integrated model the AGI. You know it does everything. It knows when to think hard. It knows when to quick. It can do all the other thing that Samman actually said was the fact that advanced voice mode is going to get an update for GBT 5. The world was pretty stunned when we saw GBT4 o voice mode. So, I'm pretty sure GPT5 voice mode is going to stun us as well. Think about something that's extremely lifelike. If you know the recent demo of Mayor, it's quite likely we'll see something like that. And then, and then of course, Solman also said that GT5 is quite likely to be smarter than any person on Earth. How many people feel smarter than GPT4? Okay. How many of you think you're still going to be smarter than GBT 5? I was expecting more hands here. Um, I don't think I'm going to be smarter than GBT5. And I don't feel sad about it because I think it just means that we'll be able to use it to do incredible things. And you know, like we want more science to get done. Uh, we want more we want to enable researchers to do things they couldn't do before. This is the history of this is like the long history of humanity. Um, it does feel a little different this time because of what this can enable. But if scientists can do things because they have like a crazy high IQ tool and they can focus more on figuring out the right questions to ask, address things quicker, do their search space faster, uh, that's just a win for all of us. So, we're thrilled to get to enable that.