NEW OpenAI o3 and o4-mini Update (FREE!) 🤯

There's a brand new chat GPT open air 03 and04 mini update. You can see this was just announced a few hours ago. Chat GPT is saying this is our smartest and most capable models to date with full tool access. So you can get access directly on chat GPT. If we click on the top left over here, you can see 04 mini high 03 and we'll be testing out in a minute as well. So, OpenAI have said today we're

releasing Open AI 03 and04 Mini, the latest in our O series of models trained to think for longer before responding. These are the smartest models we've released to date, representing the change in Chat GP's capabilities, blah blah. For the first time, the reasoning models can agentically use and combine every tool within Chat GPT. So, we'll see how that goes. So basically what this means is you can use like the web search, you can analyze uploaded files, you can take data, you can use reasoning visual. So it's multimodal reasoning basically. And then one of the biggest changes here is that it is trained to switch between tools and figure out okay which one is the best to use. Now according to the benchmarks openio3 is the most powerful reasoning model for coding, math, science, etc. and it's supposed to perform strongly at visual tasks like analyzing images, charts, and graphics, making less mistakes than OpenAI 01. Now, you might be saying, what's going on here? Because shouldn't O4 be more advanced 03? So they've said the OpenAI O4 Mini is a smaller model optimized for fast cost efficient reasoning. It achieves remarkable performance for its size and cost particularly in maths coding and visual tasks. So this is like smaller model. It's a bit lighter. It's a bit cheaper if you're using the API blah blah. You can get access to these models for free as well if you're wondering about that. If we have a look

at access here, OpenAI have said chat GBT plus pro and team users will see 03, 04 mini, and 04 mini high in the model selector starting today. And this has replaced 01, 03 mini, and 03 mini high. Now free users can actually try 04 mini by selecting think in the composer before submitting their query. And there's rate limits across all the plans. And the plans remain unchanged. Now, if you're wondering, okay, what are the benchmarks here? How does it perform?

So, you can see all the models here. So, you got 01, 03 Mini, 03, and 03 Mini, which actually seems to be performing the best. Then you've got 01, 03 Mini, 03 no tools, and 04 Mini, no Tools. So, it looks like on the benchmarks for Competition Maths and Amy, 04 Mini is performing the best so far. always take these benchmarks with a pinch of salt, but you can compare each one side by side here. Here's the same with maths. So, you can see actually for maths 03 mini is performing the best. For coding, 03 is performing the best and then 04 mini with 01 performing a long way behind. And for aentic tools, 03 seems to be performing the best by quite a long way versus 04 mini versus the others. One thing to bear in mind here is these are all evaluated at high reason and effort settings which of course would cost more money to run right. So if you're using the high reasoning effort settings on your API that would cost you more money. So these are where the benchmarks are. So just bear that in mind as well. If you tune it down a little bit then you probably won't get as good results. What's interesting here as well is it's got this new reasoning model when it comes to images. So you can see, for example, someone has said, "I've taken this picture earlier today. Can you find the name of the biggest ship you can see and where it will dock next? " Actually reasoned for 93 seconds and then came up with some ideas right here. I don't know how often you're reasoning with models, but you can do it. Now, you can also take like drawings and just upload them and then use that for thinking mode. And you can see the differences between OpenAI 01 versus OpenAI 03. So for business, science, sports, and visual reasoning directly on the comparisons

here. So if we go into chat GPT now, we'll select 03. We're just going to grab some example coding prompts from the AI profit boardroom. So let's test this out for creating a game. So I've said make me a captivating endless runner game. Key instructions on the screen. P5Js scene, no HTML. I like pixelated dinosaurs and interesting backgrounds. Now, I know for a fact if you run this through Gemini, I've done it quite a few times now, you'll get a really good output. Let's see what we get back from 03. Now, it is quite slow. If you look at the time that it takes to reason that is quite slow, but now it's working its magic. What's also cool about this is it's working with canvas as well. Whilst that's do that seems to have cut off here, but let's see how it performs. So, we'll just run this inside p5 js and see if it works. There we go. It's not bad. Not bad at all. Actually, it worked perfectly first time round. It was quite fast. I think that's a lot faster than when you're using Gemini to create the code. So, how long did that take? Let's have a look here. So, it took 9 seconds to think about it and then I would say it just took a few seconds more to get the output. You can do the same thing inside Gemini as well. So if we take this prompt, how long is that going to think for? I just want to see how much longer that takes to generate. So that's going to be thinking about it for quite a lot longer. Previously 03 was thinking about it for 9 seconds like you can see inside the chat here. Gemini thinks about it for 23 seconds and then generates the output. And I would say this is a lot slower. However, I'm super impressed with Gemini and it's free. So, not bad for its first attempt. Let's see what's happening. Sam

Alman as CEO of OpenAI actually said 03 Mini are super good at coding. So, we are releasing a new product codeex CLI to make them easier to use. This is a coding agent that runs on your computer. It is fully open source and available today. We expect it to rapidly improve. Here's a link and you can get that on GitHub. If you're wondering how to install this stuff, usually what I do just to make it super simple is I'll copy all the content on the page, use a tool like Claude and just say, "Help me install this on a Mac, right? " And it will give me the instructions on exactly how to get that set up. I might do another video on testing that and seeing how it performs. And as you can see, this was just announced 4 hours ago. So this is brand new this model. Also 03 and 04 mini are available in cursor. 03 mini is free for the time being. And another crazy piece of news is that open eye is looking to acquire Windurf for $3 billion. Wild. So they're making some people are calling it AGI. I'm not sure about that. Let's see if it's in Visual Studio Code as well. So if you go to Visual Studio, you should be able to download it. And then if we go into for example like root code, we'll just go into settings here. So I've got my open retita API key plugged in here. And if we type in 03, you can see we've got that available to code. And let's just see what it's like for coding here. So I'm going to grab a prompt from the AI profit boardroom link in the comments description if you want to get access to that. And we'll plug this into R code over here using 03. Just see how it performs. And we have an error straight off the bat. Okay, not a good start. Let's switch to OpenAI. Just use the API key directly. We'll go with 03 mini high even though that's going to cost a lot of money. That's the price I pay for you. Grab an API key from OpenAI. And then we'll go back and plug this bad boy in. Honestly, that seems to be taking ages. I don't know whether that's the settings or not. Let's have a look what we got here. Super slow, isn't it, to code with that is taking ages and making no progress. I'm sure you can tone it down in the settings if you want to, but it's coding out now. So, let's see what we get back in a sec. So, it's done the HTML. Now, let's run it. See what we got back. Um, it doesn't look good at all. If you were using Claw 3. 7 Sonet or these other tools, it would look a lot better. This is far from AGI, my friends. And that was on the high settings, right? Let's see how much that cost us. So, it cost us 0. 07. It was cheap to build, but it doesn't look good. Wouldn't recommend it. That's as basic as it gets when it comes to UI to be honest with you. So, I don't think it's worth coding with the high settings. Maybe using the Aentic tool is better, but from what I've seen, this is definitely not AGI, mate. I'll tell you that for free. So, thanks so

much for watching. If you want to get access to my community that's focused on making money and saving time with AI, feel free to get that in the AI profit boardroom link in the comments description. This comes with all my best SAPs, tips, templates, workflows, and aentic workflows as well along with a full crash course and weekly Q& As's where you can jump on and ask us any questions. We have a really active community of 670 members. So, you can post in here with any questions you have or anything that you want to share and join the community. And on top of that, if you can't make the Q& A, they actually get recorded and you can watch them back here. And if you want to get a free one-to-one SEO strategy session that shows you how we retail websites from zero to 145,000 visits this month and generate hundreds of thousands of dollars in sales on autopilot, feel free to get on this free link building acceleration session. You're going to get a free custom tailor game plan to generate more free leads, more free sales, and more profits from Google directly to your website along with other AI search engines based on what's working for our happy customers like you see right here. And I'll see you on the next one. Cheers. Bye-bye.