OpenAI has just rolled out chat GPT 5.4. There's actually a couple cool features in here that I'm really excited about, that I've been wishing CHATGP has been able to do in the past and they finally launched it. And of course, if you look at all of their marketing, it's gonna just basically be them saying, this is our most capable model yet. And of course, it's a most capable model if it wasn't, I mean, what would they even be making an update for?
So I'm just gonna get past all of the hype and all of the buzz from what they said in the launch. And I'm gonna tell you some really interesting use cases and some ways that I actually think this is useful. GPT 5.4, before we get into all of that, if you wanna try all of the latest models, go check out my startup, AIbox.ai. We have the latest models from the top 15 different AI companies, everything from Grog to Gemini to Anthropic to OpenAI, 11 Labs for audio, tons of cool image generation models.
I think there's over 50 models on the platform total. You can try all of them side by side and it's only 8.99 a month. So much cheaper than CHATGP. But you get way more models.
And of course, you can also use it to automatically create AI workflows that can complete tasks for you that are automated. So there's a ton of cool stuff going on, but go check out AIbox.ai if you wanna get access to all of the top models for only 8.99 a month. And it's 20% off if you can annual plan as well. So there's a lot of cool stuff there.
All right, let's get into what's going on. The first thing I wanna mention here is that this is called GPT 5.4 thinking. They have a higher performance variant that is known as GPT 5.4 Pro. But both of these together are designed to handle everything from some complex analysis.
They do a lot of coding, a lot of long running workflows across a lot of different professional software tools. And they're kind of dubbing this as like their professional work tool. They're trying to get into the hands of more working professionals. And this is coming right on back to them signing a whole bunch of deals with a bunch of different consulting firms that are gonna allegedly get CHATGP into more businesses and kind of the professional environment.
At the same time, they're having kind of this locked in a battle with even Googles in this right now, but really with Anthropic for Anthropic Cloud Code. They're a code X tool. They're really trying to push forward in kind of how software is using AI models and how computer use is going on. So this is where they're really focusing.
One of the biggest changes basically about this is the scale. So in the API GPT 5.4 has a context window about to a million tokens, which basically lets them work with huge documents, really big conversations, big data sets. And really, I mean, if you think about this, a huge benefit is going to be coding where you can look at bigger code bases to actually work with. So something that's on the Anthropic was really crushing and then opening eyes trying to get into this.
Opening also says that the model is specifically more with the same as token efficient, which this is actually one thing that I'm excited about, basically can solve the same problems using a lot less tokens than GPT 5.2. So your costs are going to come down. It's actually kind of cool. If you already had 5.2 running in a software, which even if you don't, a lot of the software you use will, the costs come down a lot for that.
And it also gets a lot faster. So the cost come down and the speed goes up. And so yeah, for me, this is something I'm actually excited about. So as far as how the benchmarks look, I know, you know, I'm not trying to like sit here and pick the benchmark percentages, but I did want to talk about some interesting use cases and reasons why these are why they're good.
Specifically, it's kind of leaning on a bunch of the better known benchmarks. One of those is for coding. Of course, we know why that's important right now, but also computer use. And this is something I'm excited about right now.
I feel like anthropic is really crashing with computer use. Basically, you know, it can look at everything on your computer and go click on stuff and get stuff done for you. This is a use case that I've been using a lot with Clods and Thoropic browser, the Clod Crome browser extension. Basically, it's a button you click.
It opens a side chat bar. I got a really complex UI or complex website. I'm not a developer, but if I'm going into like, for example, recently I had to do some stuff on Google Cloud to set up a tool that I was building on Lovable, and I needed to beef up my back end so I could do some extra fancy stuff. I didn't really understand anything that Lovable was telling me I needed to be able to do.
So I opened up the Clod sidebar, told it, look, I'm on my Google Cloud account, and here's the instructions for Lovable, and it clicked around and set up some stuff for me. Now, should I have a real developer look over this? I mean, we're going to throw caution to the wind for the time being, and I hear all the developers screaming into their headphones right now. But at the end of the day, it got it down and my software is now functioning.
And I did not have to watch a whole bunch of long YouTube tutorials on how to set up some complex, I mean, for me, complex, because I have no idea how to code, Google Cloud stuff. So this is a really incredible use case for a lot of reasons. And I think opening up their capabilities and computer use is really exciting because they're going to start competing more directly with open AI. I mean, this is not like Arabic.
It's not like anthropic. It's kind of like the only one working on this. Open AI has been doing this for a long time with agents, but it feels like it's getting a lot better. The other one that I'm excited for is they're getting a lot better at knowledge work.
And so these are kind of things that I think everybody uses it for. So this is something we're just going to see some incremental improvements on. Open AI's GDP Val benchmark, which basically checks tasks. It has like 44 different occupations.
So it's kind of like showing you how you can use this for different professionals. It is exceeding industry professionals in 83% of comparisons. So they're like, look, these are all the tasks that people in all of these different professional industries are doing. It is better than 83% or it's being one of industry professionals might give you in 83% of these cases, specifically, I think for knowledge work.
And it has a really big jump from achieving about 71% to GPT 5.2 is getting. So upgrading this to now GPT 5.4, we're getting from 71%. It's just basically is going to be a lot better for knowledge work. Which is pretty significant.
On some of the coding benchmarks, so SWE Bench Pro, this is a software engineering bench pro, the model is getting slightly better than the last version. So I mean, this is good. But beyond just getting slightly better, it is actually a bit quite a bit faster. So if anybody has used a lot of these software tools, specifically, we use Cloud Code AI box.
And my developer sends me screenshots of these really long elaborate tasks that it's doing on our back and our code base. And I swear, it's like a goal for him to see how long he can get Cloud Code to run continuously without stopping on project. He gives it. It's funny because I'm live coding stuff on lovable.
And I usually get a lovable response back to me in a minute or two. He has a goal for like three and a half hours doing a task. So in this model, it gets faster. I'm excited because hopefully that three and a half hours gets cut down on some of the stuff that we're working on.
I think one of the things that it's also very good at is for real computer interaction. There's an OS world verified. It basically evaluates how well an AI can operate a desktop environment. It's pretty much just like takes a screenshot.
And then it uses the keyboard and mouse commands to go and click stuff. Right now, it has about a 75% success rate. I've used Cetropy agents. It's not perfect.
It's actually not my go to. I don't use it that much. I wish I could use it more. I think Mthropic is doing better in this.
75% success rate, like they are improving their success rate is up a bit. It's better than GPT 5.2. I still don't think it's the best. There's a major focus on kind of how it is being used professionally.
Open ISAs, their model right now is significantly better at basically giving the kind of deliverables that people use in real work. So things like spreadsheets, presentations, financial models, legal analysis, all of those, they've done a bunch of different tasks. And they had one performed by a junior investment banker analyst. It got 87% compared to 68% that GPT 5.2 got.
Some human evaluators also preferred it about 68% of the time. They said it had better visuals and better structure. So there's some cool stuff. OK.
Cool features that you might actually use today. This is the one I'm very excited about. It has what they're calling steerability. But basically, when you're talking to chat GPT, it's available in the API, too, which is crazy.
But it's on chat GPT. If you're talking to chat GPT, and you can kind of see it's reasoning, it's thinking through some stuff, and it puts a couple of steps down. You realize it's going the wrong direction? Maybe you're like, hey, I'm trying to visit the best beach for surfing.
And it's like, OK, looking at beaches and Kauai, and you're like, oh crap, I'm in California. I don't want to see Kauai. And you're like, then you can type a message, specifically in California. And mid-prompt, mid-response, it actually takes into account what you just said and is steerability.
It's going to go and incorporate that into what it's looking at, and into its reasoning, and give you an updated response. So basically, you can do mid-response prompts, and it's going to take that into account, and change its prompt, and give you a better prompt, mid-response. So that's kind of interesting, because I think they did a couple of clever things here. But one of them is like, when you ask a question, you have to wait for it to think, you have to wait for it to reply.
You sit there, and you wait. We all hate waiting. And so if in the middle of waiting, we're reading it's line of reasoning, and we're giving it more input and more feedback, it feels like we did a lot less waiting. So you have to wait for it to spin up the whole thing, and you'll be like, OK, this is wrong, and here's why it's wrong, and here's what you should do instead.
You can do that in the middle of the chat conversation response, which is really cool, in my opinion. Something else I've kind of focused on right now is online research. Apparently, it can search across a greater number of sources on the web, so it can kind of instead of just like, OK, we're looking at this website, get in some data, and we're going to look at this website. It's going to go in and search just like a ton at the same time across the web.
And then it's going to follow leads across different pages. So it might get an idea from something to read on one article. It's going to go follow that to another article, and then it's going to go around a lot more. So it's kind of doing, I know we have had deep research for a while, but it's doing deeper research if that's the thing.
And it's going to combine all the information that it gets into one coherent answer. So basically, it's going to be more useful for some of the more complex questions, where information is kind of scattered across a lot of different sites instead of sitting in one place. Now, not every question you ask is going to be relevant, but sometimes when you have a complex question, it's going to be able to go get you a more coherent answer quicker. So this is great.
They have all this kind of fluff in their launch about hallucinates less, and it has less factual errors and all this kind of stuff. I don't think that's super important. One thing that we also heard about is that it's like, hey, you ask a question. It's less likely allegedly according to that moment to not answer.
However, our good friend Connor Grannon, who hosts the AI Applied Podcast with myself, he was testing it. I saw a post he made on LinkedIn where he asked it, is it true that air bubble inside of an IV can cause me or could kill me? And it said, apparently it typed out the whole response to him, kind of, and just like we saw with deep-seak and the Chinese-sensor model, if you ask anything about Tiananmen Square to deep-seak it, it types it out, and it disappears. It's like, sorry, can I?
Apparently, Cheych is the exact same thing. And also, this is kind of a tricky moment because we're seeing New York right now is trying to pass the legislation, where they're saying, hey, they're basically trying to pass legislation, saying AI models can't answer any questions about medical, health, legal, like they have all of these different areas. I think even hairstylists, they're trying to put in areas basically all of the different industries of regulatory capture, they just don't want people to be able to get the answers for free. So pretty, I don't know, kind of bummed about that legislation and people seriously considering that.
However, so it doesn't seem like it's that much better, but maybe it's moving in a good direction on not 100%. Sure, it still feels like there's other models that are more of the adult in the room. But you also get pros and cons with those models. Grog famously is going to answer any question you have about basically any of those topics.
But there might be some other cons with Grog. So pros and cons to all of the models. Thank you so much for tuning into the podcast today, guys. If you enjoyed the episode, it would really help the show a ton if you left it a rating review wherever you listen to your podcast.
Just drop me a note, say if you enjoyed it, say where you're from, say what topics are interesting to you. I read all the reviews and all the comments, it helps a ton. Also, make sure you go check out aibox.ai. If you want to get access to all of these latest models in one place, you don't have to pay $20 subscription to 10 different platforms, it's $89 a month and you get access to over 40 different AM models.
So go check it out, link in the description, aibox.ai. I'll catch you guys all in the next episode.