03, 04, 01, GPT4.5, GPT4. There are just too many models to choose from. In today's video, I'll be explaining to you every single OpenAI model, what they are, and when to use them. So, without wasting any more time, let's get on to the first model. So, let's start with the smallest model. Now, if you scroll down and you go over to here, you'll actually see this model hidden away. This is GBT40 Mini. Now, what GBT40 mini is, this is the fastest one that is basically for everyday tasks. What everyday tasks, you might ask? Well, think about the quick tasks that you have on a day-to-day basis that don't really require much sort, but you just need a little bit extra nudge in terms of thinking that could help your decision-m. Now, one of the things that you need to know about GBT40 Mini is that it's likely that you'll never get rate limited, meaning that you could probably message this thing all day, every day, and your chats will stay there running smoothly. So, GPT40 mini is a compact, costefficient version of OpenAI's GPT40 model designed to deliver powerful AI capabilities at a fraction of the computational cost and the latency of larger models. Its primary use cases center around making advanced AI accessible for a wide range of applications where speed, affordability, and scalability are critical. So GPT4 Mini, it's basically a faster, more compact, efficient version of GPT4. Now, one of the main use cases for GPT4 Mini is customer support automation. Most people that use this model, they essentially build this into chat bots and virtual assistants that handle large volumes of customer queries. Remember, those queries need to be fast, high quality, and in real time. For the average person, this could basically be customer queries about your entire life. And let's say for example, I need a very fast query that doesn't require that much deep reasoning. I could ask something very simple such as, "Do eggs in the UK need to be in the fridge or not?" Now, sometimes it will actually search the internet, but since it is such a small model, finding quick facts that can help you on a day-to-day basis, this is exactly what the model is basically for. You can see right here, it gives me a list of different pieces of information, and it was able to research that really, really quickly considering that this was a smaller model. You'll notice that this one is much, much faster than the other models. And this is why you want to be using this model when you have those quick questions that can be answered within seconds that don't require any deep reasoning. Of course, if you were maybe even baking something or wanting an entire recipe, you'd probably not want to use this model, but this is something that you just need for really quick questions. Now, with the GPT40 Mini, you can input things like images, but it's rather that you don't, considering that this model is basically just a model that gives you a quick response back when you need something really quickly. Next, let's take a look at GPT40. This is a model that doesn't actually excel anywhere, but it's actually pretty good for a wide variety of tasks. Just like GBT40 Mini, this is a chatbot that can basically do everything and anything. But like I said, it doesn't really excel at one specific feature. It's more of an all round chatbot that you would use if you were getting work done. So let's say for example, maybe you needed to write an essay or maybe you needed to write an outline for a video script. This is the model that you would use. And the reason I like this model is because it's trained to explain things and format things in a way that's visually appealing. It doesn't necessarily output large blocks of text that really confuse you. It really just outputs text that is a really easy to digest. For example, I'm saying, "Write me a video script that explains quantum computing to 10-year-olds." So, I'm going to input this in right here. And you can see immediately it starts to give me back this information that has this video script and all of this information. Now, I did mess up my prompt here a little bit because this is not really the kind of video script that I do want. But it does actually give me information relatively quickly in a way that is easily digestible. And since this is a video script, you can see here that it actually has the host, the on-screen visuals, and other interesting things. If we ask the same question, but if we ask it in a different way, for example, saying, "Explain this to someone with an ATIQ." We can see here that the format is also really nice. the way how it lays out the information. It manages to put certain things in bold and space it out in certain ways. It allows you to easily understand complex topics, which is what I use this model for on a day-to-day basis. Now, I don't use this model to write many things, but I more use this as a conversationalist. If you want to go back and forth with ideas, this is basically the model that you want to use. For example, what this model can do really well is it can also generate emails really well. So, if you want to generate a really detailed email, for example, let's say I wanted to leave a job, you can actually ask it to write an email and it'll actually give you an email that is really, really nice and really well thought out. This is probably the perfect model to use for this kind of task because it's able to reason in a way that doesn't require that much detail. For example, right here, you can see they are currently testing different ways to use the model. And I do actually prefer this response right here. You can see it's able to give you a decent amount of information relatively quickly. Some other things that GPT40 can do is it can actually translate text very well. So for example, I could say translate this text to Spanish and it's able to translate that really really well in blistering speed. Now this model is also very good at digesting images. Most people don't realize that GPT40 can actually analyze images on a day-to-day basis. Of course, now it's not the number one model, but if you need to quickly analyze an image and quickly dissect exactly what's going on, this is going to be the model that you quickly use. For example, I can grab this image right here, and then I can click explain this, and it's going to quickly use its multimodal capabilities to be able to dissect exactly what's in the image. I didn't even spell explain, right? But you can see right here, it's able to explain this image really, really quickly in a way that allows you to understand exactly what's going on. This model does have exceptional vision capabilities and most people don't actually realize this because they're not using it on a day-to-day basis. And once again, we can see that the way how it formats the text allows you to easily understand anything. Now, remember this model is the base model for the voice model. So, if I do click this voice mode here, you can see that it's going to pop up with this voice mode. And this is something that you ideally want to use once again when you're going back and forth with ideas or when you need to have a conversation with someone about something and you don't really want to call someone and you just really want to explore the different possibilities from an AI that knows a million more things than you do. So, hey Chat GBT, can you hear me loud and clear? What's up? Can you actually go on the internet and let me know what the latest football match for Arsenal was? I'd love to know how they're performing recently. Arsenal's most recent match was a 1- nil loss to Paris Sanjaan in the Champions League semi-final first leg. Usman Dembele scored early for PSG. So we can see here that this is a model that is of course the base model for the voice. And what's really powerful about this is that you can actually use this to quickly search things on the internet, go back and forth with the business idea, explain things that are going on in your life, or just generally have a second person's opinion on something that you might be tinkering on. And with the advanced voice mode on GPT40, this is something that is remarkably useful for the average everyday person. Oftent times I'm trying to understand complex concepts and I don't really want to type. My hands might be busy. I might be cleaning up and I'm able to just put some headphones in, talk to GP4 quickly, and I'm able to multitask in a way that's rather efficient. Remember, it does now have access to the internet. So, you can even ask it to quickly browse something and it's able to tell you and return that search result. A really underrated feature that most people don't take advantage of. Now, the next model we're going to be exploring is GBT 4.5. Now, this is a model that you need to use when you're looking at writing or exploring certain creative ideas. This is because GBT 4.5 excels at one specific area where other models don't. GBT 4.5 demonstrates improved emotional intelligence, adapting tone and language to the context of the conversation, and it actually produces more natural, empathetic, and polished responses, making it well suited for emails, professional communication, and human interaction. GBT4.5 is essentially a model that is really good when it comes to trying to persuade humans or writing texts that could be salesy. So, if you don't know, GT4 is basically the largest model that contains a lot of the human nuances that really makes this model good in ways that most people don't realize. Now, quickly before I demonstrate the model, it should be noted that you don't get too many uses of the model. The model currently has, according to many different sources, only 50 messages a week. This is because this model is by far the most expensive model that OpenI have produced. But let's try it out anyways. So, for example, what I use GPT 4.5 for is I'll use it to do essentially creative writing that needs a human touch. Now, GPT 4.5 is noticeably a little bit slower as well. So, if you need something really quickly, please don't use GPT 4.5 because it outputs text that is a lot slower than other ones. So, here I'm asking it to write a text about AGI. And you can see it's writing relatively slowly compared to the other models. Now, you can see right here, it's actually explaining the importance of AGI in a way that humans might understand. Now, one of the problems that GBT4.5 has, which isn't really a problem, is that you won't really be able to tell the difference as a human. But when they tested GBT4.5 in many various tests, it was almost as twice as impressive as GPT 4 in terms of actually acting as a human and writing creative stories and subtle nuances. So, this is a really important part of the model that you really don't want to miss. So you can see right here it says, "We have a unique responsibility at this pivotal moment in history to ensure AGI becomes humanity's greatest ally rather than its most powerful threat. The stakes have never been higher. The future of humanity depends on what we do next, which is really nice. So for this one, I would say you could use it for drafting social media post in a human tone, creative writing that maybe needs a human touch. And if you do need extra accuracy, apparently this model has a lot less hallucinations. So this is a model that is really good when it comes to factual accuracy. Now one last thing about GPT4.5 that you could do is that I would say it also deals with human emotions a lot better than any other model because this is the most humanlike model. It can really understand what makes you tick and really give you some insightful advice that may actually help you. I'm not saying that you shouldn't seek out a licensed therapist, but if you do have access to a free tool, it does make sense to actually see if the model has any valuable insights. Next, we're going to move on to 03. Now, 03 is a model that most people misunderstand. It's probably the best of the best when it comes to AI right now. And that's because it's actually an agent wrapped in a chatbot. What I mean by that is that 03 is essentially an AI agent that can go out and do tasks for you, come back to you, and give you a research report. It can do graphs. It can do so many different things that I'm actually going to have to spend a little bit more time explaining to you just how powerful this model is. No, new model is here. Now, one of the most insane things that 03 can do is truly thoroughly analyze images. Not like before, which was essentially visual classification of images. The type of image classification here I'm talking about is really get granular with the data. So, if we take a vague image of any location, and I'm actually going to grab an image from Geogesser here, and I'm going to say guess this location. Where is this? If we put it into 03, the way in which it analyzes the image is completely different. It'll start to zoom into the image. It'll start to crop certain parts of the image out. It will then start to reason with the internet and research certain things within that image to ensure it's able to classify exactly what is there. And this makes it one of the most powerful tools when it comes to analyzing images that need extra detail. So, if you are reasoning about something that needs real real detail in the real world, it'll do that in a way that we just haven't seen before. And most people don't even realize this feature exists, which means that you need to be taking advantage of this. Right here, you can see that has cropped the image in certain ways. It starts to look around. It starts to, you know, look at the background showing different things. And we can see right here that this is one of the most important features of the model that other ones don't have. 03 here is essentially using tools to achieve its tasks. And one of those tools is the ability to crop the image and also do a Google search. So it's able to cross reference that with many different sources, view the internet, and then it's able to come back with a final answer. Right here, we can see that it's actually searching the web because it managed to find some details and it wants to double check to see if it actually looks the same. You can see right here that it says to verify the walkway's vantage point actually search for images of this state park and confirm that it indeed does have the same walkway. So, the location it managed to give me says that Liberty State Park in New Jersey, you're standing on the Liberty Walk from Prominade just south of the US flag plaza with the Jersey City skyline across upper across upper New York Bay on the far right. And so if I do this on geogesser, I'm actually going to put the pinpoint right here. And if I click guess, let's see how close it was. It was literally extraordinarily close to the right guess. So that is the power of 03. I mean, that kind of guess is absolutely incredible after just taking a random picture of something. Now, another good feature of 03 is essentially advanced business analysis. Because this model can actually go out on the web and do a different things and use different tools, it's able to analyze your business in a way that we've never really seen before. So, I'm actually going to swap to a recent video I made where I spoke about the fact that this can analyze businesses and data and give you visualizations using the tools in a way that helps you out that is absolutely tremendous. It's able to forecast insights with remarkable accuracy. And because it's able to browse the internet, it's also able to ground that in real world data. So overall, you can see here that 03 is by far the most powerful AI. It's able to reason through different things, use different tools, browse the internet, and think through a problem in a way that we really haven't seen from any other models, which makes it probably the most powerful model that you want to be using for complex tasks in any kind of STEM domain. Now, let's move on to the other models, 04. Now, 04 mini, I'm actually going to batch this in with 04 mini high. Essentially what this model does, this model excels at mathematical problems. Now most people on a day-to-day basis aren't really, you know, tackling complex mathematical problems, but if you ever do have something that is highly based in mathematics, maybe you have a business calculation that you need to work out, maybe you really want to figure out how much profit you're going to want to make in a business quarter, 04 is the way to go. It's currently number one when it comes to really difficult prompts in math. And I think that it's quite underrated considering that most people don't really have a use case for it. Now, in order to generate synthetic data for this question, I've actually asked the standard model GPT40, which is really good at synthetic data generation in terms of basic things such as generating basic data for a business that you want to virtually test. It can easily output things in the CSV format, which allows you to import it directly into tables. So, here I'm going to go to O4 Mini. Then I'm going to add this data. And I've added this spreadsheet and I'm going to just input this question here. It basically says, assuming each order contains only one item and the cost of goods sold per item is $60 and the fixed monthly operating cost of 50K, calculate the total net profit over the six-month period and then determine yada yada yada. So here you can see 04 mini is going to think and reason about this problem in depth. Considering I've added data here, I know it's not that much data. This is a model that is really able to once again use tools and is able to excel in the math domain. They've trained this model using reinforcement learning to really pursue good chains of thought when it comes to thinking in the right way about how to solve the math problem. So you can see right here it's able to give me a quick solution and it's quite likely that this is correct. Other models really don't excel at mathematics, but this is truly where 04 and 04 mini high really do win. So if you do have anything that is highly mathematical based, this is where you'd ideally want to use 04. Now I also do want to add one thing that sometime gets people confused. There is essentially two forms of the same thing. You can actually ask 03 to create an in-depth research report for you or you can use this button right here that is basically giving you an in-depth research report. Now with deep research and 03, OpenAI actually showed us in the benchmarks that they're practically the same in terms of quality. So realistically, if you want, you could really use deep research or 03 for the same query. The only difference that I would say is that deep research is basically just going to research a topic quickly for you. Maybe if you're writing an essay, that's where you'd want to use it. But with 03, this is more about an agentic solution. So if you wanted it to potentially put that data into a table or think about the solution to that problem, that's where you'd want to use that over the deep research feature. So for here, I'm asking it for a research report on the AI trends predicted for 2026, the top 15. So the difference is once again, what this will do is it will actually ask you some follow-up questions so that it can do its research in depth. And it's quite likely that this research report will take between 5 to 15 minutes to complete. Whereas with 03 will likely give you a response in under 3 minutes. The only thing is that remember deep research, you usually have more queries, but with 03 you only have around 50 a week or around 100 a week. So those queries since the model is using a lot more tools tends to take up a lot more data. Now we can see here that this has already been searching for around 5 to 6 minutes and it probably will take another 5 to 6 minutes to return the full research report. Of course if you want to open up the activity you can just hit the tab and you can see exactly what it's done. And I would actually recommend doing this if the research report isn't exactly what you want. Oftent times I'll include a step-by-step guide on where it needs to research, how it needs to put that data forward. That way I know how the final source came to be. Of course, I do this for very specific scenarios, but for this one, it was just extremely vague to demonstrate the difference between 03 and the deep research. And so, yeah, it's literally been around 25 minutes, and this deep research tool has yet to finish. So, I would even argue that potentially you could do this at the end of a workday and come back in the morning because it definitely interrupts your workflow just a little bit. You know, you can put this to work and then go about and browse.