AI transcript
0:00:03 Hey, welcome to the Next Wave Podcast.
0:00:04 I’m Matt Wolfe.
0:00:06 I’m here with Nathan Lenz.
0:00:12 And today we’re going to try to do our best to demystify all of these recent launches with
0:00:19 OpenAI because they’ve released 4.5 and then 4.1 and then 04 and then 03.
0:00:21 And it’s just getting really, really confusing.
0:00:27 So this episode is designed to break it all down for you, tell you which each of these
0:00:32 models are good for, what they’re not good for, and we’ll even rank them from dumbest
0:00:35 to smartest at the end and tell you which ones to use for which things.
0:00:40 So without further ado, let’s just go ahead and jump in and talk about all these recent
0:00:40 OpenAI announcements.
0:00:47 I think probably the best place to start is with this one right here, because this is probably
0:00:52 the first piece of news in this sort of sprint that OpenAI is on of releasing news, which
0:00:56 is that they just rolled out this memory feature in OpenAI.
0:01:02 And to me, this is probably one of the coolest, most valuable features they’ve rolled out.
0:01:04 I found it really useful.
0:01:04 Yeah.
0:01:08 I know Nathan, before we hit record, you were saying something to the same effect of like,
0:01:10 out of everything, the memory is probably the coolest.
0:01:11 Yeah.
0:01:15 So it’s funny, the team at Grok, they reached out to me about what should they build, like
0:01:16 maybe like two months ago.
0:01:18 And I was like, focus on memory.
0:01:23 Nailed that first, because OpenAI has like a half-baked version of memory.
0:01:26 They’ll probably eventually get there, but who knows if it’s a priority or not.
0:01:28 If it’s not, you should get there first.
0:01:29 Right, right.
0:01:33 Unfortunately for Elon Musk and Grok, OpenAI definitely got there first.
0:01:35 And I think it’s going to be one of the biggest things long term.
0:01:38 It’s the kind of thing where people don’t realize why it’s important now.
0:01:40 But over time, it’s going to be clear why it’s important.
0:01:44 You know, I use ChatGPT a lot, like for so many different things, like personal stuff,
0:01:45 business, everything.
0:01:48 I use it for so many different things, probably more than the average person.
0:01:50 And I asked it to like, explain what do you know about me?
0:01:52 And I’m not going to show you what it said.
0:01:54 Something that was so revealing.
0:01:55 Was it not nice?
0:01:56 No, it was nice.
0:01:57 It was nice.
0:01:59 You know, because it knows a lot about my life too.
0:02:00 My life is pretty crazy.
0:02:05 And just like some of the insights that drew out of that were just kind of shocking and inspiring.
0:02:08 Honestly, it was like, yeah, I have done a lot of great things.
0:02:10 I have went through a lot of hard times in my life and all this stuff.
0:02:14 And it was just, it seemed to know things about what I want to accomplish in my life.
0:02:18 Things that I don’t think I’ve told it that were just kind of like, that’s me.
0:02:24 And so the idea that AIs are already starting to actually understand me maybe better than I understand myself.
0:02:24 Yeah.
0:02:26 And I mean, you’ve probably had a lot of chats.
0:02:33 I know you’ve mentioned that you’ve done even like voice chats where you go for like walks and just sort of like unload your thoughts into the voice chat.
0:02:35 And I mean, all of that is in memory now.
0:02:38 I mean, and just to be clear, you can turn off the memory.
0:02:41 If this feature freaks you out, there is an option to turn it off.
0:02:49 So like you don’t have to let it remember everything, but like, you know, all of those voice conversations where you’re just sort of like brain dumping.
0:02:53 That’s all in its context now when you chat with it.
0:02:54 Yeah.
0:02:57 I think long-term that’s how AI is going to get really useful for people.
0:03:04 Cause like people think of it right now, it’s just chat to you where you go in there and you copy and paste something in, maybe have it help you in some way, uh, or you just chat with it.
0:03:11 But eventually these models are going to understand maybe what you want or what you want to accomplish better than you yourself understand that.
0:03:13 Or even before you realize you want it.
0:03:14 Or before you do.
0:03:14 Right.
0:03:16 And this is like, this used to be like the dream.
0:03:19 Like when I first got involved in Silicon Valley, like we were all dreaming of this stuff.
0:03:21 We’re like, how do we make the web better?
0:03:23 Like, well, you got to make it more personalized while we actually understand people.
0:03:25 It’s like, how do you do that?
0:03:27 I guess machine learning or something.
0:03:28 And like, we didn’t know how to actually do it.
0:03:33 And people started doing it with like a very basic version of like, oh, well, Amazon’s kind of pulled off something like that.
0:03:36 Cause they kind of get what you want, but it’s not perfect.
0:03:37 Same as Netflix, not perfect.
0:03:42 Eventually all of this stuff, it’s just going to know about the kind of stuff you want to accomplish or do in your life.
0:03:46 And I think that’s where we’re going to get to a point where AI is going to be really helpful.
0:03:49 Anything you’re wanting to do, maybe it’s going to recommend it to you.
0:03:59 Beforehand, like you said, it’s going to know you or it’s going to just know the kind of stuff that you want to do and be able to help you better with more personalized help versus just some generic stuff that it says to everyone.
0:03:59 Yeah.
0:04:08 I mean, sort of random side topic, but since you mentioned Netflix, apparently Netflix is also going to be using OpenAI’s AI tech inside there.
0:04:12 So it will actually even better understand you and what movies to recommend.
0:04:13 But that makes sense.
0:04:17 Like I said, when I first got involved in Silicon Valley, everyone was trying to figure out how do you do that?
0:04:23 Cause it felt like the natural evolution of software products was to understand users and what they wanted better on a personal level.
0:04:25 And no one really nailed it.
0:04:33 Like I said, Amazon and Netflix got somewhat close and everyone’s just been like tweaking on what Amazon Netflix has been doing for like 10, 20 years now or something like that.
0:04:34 So yeah, it makes sense.
0:04:34 Yeah.
0:04:41 Well, I actually like the very first thing when this new memory feature rolled out was there was a button that says, do you want to try it?
0:04:42 And I said, yeah, sure.
0:04:42 Let’s try it.
0:04:45 And it automatically pasted in this prompt.
0:04:46 Describe me based on all our chats.
0:04:47 Make it catchy.
0:04:49 This is what it said about me.
0:04:50 I don’t mind sharing mine.
0:04:58 It says you’re Matt Wolf, a future forward creator, a high trailblazer, and the go-to tech Sherpa for over a half a million subscribers.
0:05:00 You’re not just riding the wave of innovation.
0:05:02 You’re helping build the surfboard.
0:05:07 Your YouTube channel is a hub for creators, entrepreneurs, and curious minds hungry for what’s next in AI and tech.
0:05:14 You break down the complex, spotlight the useful, and always bring the human angle all while rocking a retro futuristic vibe.
0:05:21 Whether it’s a short, punchy, real, or deep dive 20-minute banger, your goal is always the same.
0:05:24 Empower your audience to do more, create faster, and stay ahead.
0:05:29 Off camera, you’re a family guy with a white pup named Bowie, a deep love for camping,
0:05:32 and a soft spot for the San Diego Padres.
0:05:35 You’ve got a musician’s soul, a builder’s brain, and a storyteller’s heart.
0:05:39 Even when burnout hits or the algorithm plays rough, you keep showing up.
0:05:41 Driven by purpose, not just metrics.
0:05:43 You’re not just documenting the future.
0:05:45 You are part of shaping it.
0:05:48 I was pretty flattered by that, so I don’t mind sharing it.
0:05:49 Yeah, yeah.
0:05:53 I mean, like, parts of that were kind of generic, and then parts of it seemed like it was, you know, kind of personalized view.
0:06:00 I have a feeling I probably talked about more personal stuff with it because the stuff it said to me felt more personal than that.
0:06:00 Yeah.
0:06:06 And I mean, I also even gave it, like, a primer in the custom instructions, like, here’s who I am, and here’s, like, a general outline of my life.
0:06:08 Yeah, yours was probably like, and you keep refusing to get that wart checked out.
0:06:09 Go get it.
0:06:19 So, you know, I probably provided it a lot more about myself, but it said a lot of things, you know, it’s just, like, reading between the lines, but it said a lot of things that I did not tell it, though, like, exactly.
0:06:29 Like, it, like, kind of, like, understood things about me, about how I feel about myself, about my accomplishments, my failures, things I want to accomplish before I die, like, all these kind of things.
0:06:30 It was just kind of shocking.
0:06:32 And it was stuff you didn’t actually tell it?
0:06:34 It sort of figured it out?
0:06:34 Some of it.
0:06:35 Some of it.
0:06:39 And I was like, holy, you know, it’s, like, it’s connecting the dots.
0:06:45 It’s, like, understood my personality somewhat and the things that I get inspired by or that I want to do or that bother me or whatever.
0:06:45 Right.
0:06:52 But it definitely connected some dots and said some things that I did not explicitly tell it, which was just, holy crap, that’s amazing.
0:06:53 Yeah, yeah.
0:06:59 For me, this one feels, like, pretty on par with, like, the various discussions I’ve had, right?
0:07:02 I have, you know, talked about my YouTube channel a lot, right?
0:07:06 I use it a lot for, hey, I need an idea for this YouTube video.
0:07:07 Here’s the transcript.
0:07:08 Help me come up with a title.
0:07:09 All right.
0:07:11 Help me come up with some concepts for thumbnails, right?
0:07:14 So I go to it a lot for, like, ideation around my YouTube channel.
0:07:18 So it obviously knows a lot about the fact that I do stuff on YouTube.
0:07:18 Yeah.
0:07:25 It has all the transcripts of my YouTube videos because I always copy and paste them in and go, hey, how should I title this thing?
0:07:25 Right.
0:07:30 So a lot of this is based on probably transcripts from my YouTube videos.
0:07:30 Right.
0:07:36 It’s kind of random that, like, the one sort of personal thing it picked out about me is that I have a dog named Bowie.
0:07:36 Right.
0:07:40 Not that I’m, like, married with two kids or, you know, things like that.
0:07:41 Right.
0:07:43 But just I have a white dog named Bowie.
0:07:46 I’m like, I don’t actually remember telling you that, but it’s true.
0:07:49 So I’m sure it was in, like, one of my videos or something somewhere.
0:07:49 Right.
0:07:50 Yeah.
0:07:56 I mean, mine was more personal and, like, to the point where I didn’t cry, but it made me feel emotional, like, reading it.
0:07:57 And I was like, that is crazy.
0:07:58 And just imagine where that goes.
0:08:00 Like, we always say that’s the worst it’s ever going to be.
0:08:00 Right.
0:08:01 Right.
0:08:01 Right.
0:08:07 And the more you use it, the more it’s going to know about you and the better response you’re going to get on things like this.
0:08:11 We’re totally going to have the thing that, like, you know, in the future, when you pass away, you’ll have, like, a memory of who you were.
0:08:17 It won’t be perfect, but, God, it’ll probably be, like, a 90% of the essence of, like, who you were as a person.
0:08:21 And that is just shocking that, like, that technology is, like, basically, it’s almost there.
0:08:22 It’s, like, it’s right there.
0:08:24 It’s getting really close.
0:08:27 And, you know, this was sort of just the first announcement.
0:08:29 This announcement came out on April 10th.
0:08:29 Yeah.
0:08:30 Right.
0:08:33 So this was sort of, like, the first in a series of, like, almost every day.
0:08:33 Yeah.
0:08:36 The most important thing Grok could have built, they beat them to it.
0:08:37 Small little announcement.
0:08:39 And then on to the real things.
0:08:39 Yeah.
0:08:39 Yeah.
0:08:42 They didn’t even seem to make much of a big deal about it.
0:08:44 I don’t think, like, they even did a sort of live stream for this one.
0:08:47 Almost everything OpenAI does a live stream about in this one.
0:08:48 They didn’t even.
0:08:50 They’re just like, here, here’s a new feature that we rolled out.
0:08:53 Well, they’re talking about doing a social network and stuff now.
0:08:57 Which, if they do that, memory is a core feature of that of actually understanding who the people are.
0:08:57 Right?
0:08:59 That’s how you would build some kind of new social network.
0:09:03 Because it’d be highly customized based on the personality of the people in it.
0:09:03 Yeah.
0:09:07 I mean, I actually haven’t read too much into the whole social network thing.
0:09:10 I just pulled up an article from Reuters about it.
0:09:14 But, yeah, that’s another announcement that has sort of bubbled up.
0:09:19 Is that OpenAI is working on an X-like social media network, which is interesting, right?
0:09:25 Because there’s, like, a pretty big feud between Sam Altman and Elon Musk.
0:09:25 Yeah.
0:09:26 But it makes sense, though.
0:09:28 This is why I told the Gronk guys, like, that’s why they should do it.
0:09:30 It’s because they already have all that data.
0:09:30 Yeah.
0:09:32 About what people are sharing and how they’re interacting with people.
0:09:36 And so that would be their advantage over OpenAI is that data.
0:09:36 Yeah.
0:09:37 So what makes this one different?
0:09:39 Have you read into this one at all?
0:09:40 About what OpenAI is going to do?
0:09:41 Yeah.
0:09:43 Like, why would people use this over something like X?
0:09:45 It seems like it’s all rumors at this point.
0:09:46 I’m not sure if it’s actually even confirmed.
0:09:47 Yeah.
0:09:50 So it looks like it’s all sort of speculation at the moment.
0:09:50 Yeah.
0:09:51 Speculation.
0:09:53 I would assume that they wouldn’t be going after X.
0:09:55 They would be going after Facebook, if I had to guess.
0:09:59 Because Facebook is that kind of product where it’s all about, like, all your friends,
0:10:01 all these connections that you have, like, real life.
0:10:05 And I feel like OpenAI would be in a great spot to, like, revolutionize that kind of product
0:10:07 versus, like, competing with X.
0:10:08 Yeah.
0:10:08 I’m not sure.
0:10:10 I think there’s not a whole lot of details yet.
0:10:11 Yeah.
0:10:11 Super interesting.
0:10:14 We’ll kind of keep an eye on how that one plays out.
0:10:15 I’m sure Mark is, like, we’re super happy.
0:10:17 Like, his AI effort is totally floundering.
0:10:20 And then now Sam Altman may be coming to kill him.
0:10:22 I think it’s more likely to kill Facebook than X.
0:10:24 So, yeah, that’s a tough spot to be in.
0:10:24 Yeah, yeah.
0:10:29 So one of the other things that OpenAI has been doing is they’ve been releasing these new models,
0:10:31 but they’re also sunsetting a bunch of other models, right?
0:10:37 So apparently GPT-4 is going to be phased out on April 30th.
0:10:44 And then also GPT-4.5, which up until this week was, like, the newest model they released,
0:10:46 is also going to get phased out.
0:10:51 They sent out an email to all of the OpenAI API users, or maybe it was everybody in ChatGPD.
0:10:52 I don’t know.
0:10:59 But they sent out an email basically saying, we’re going to phase out 4.5 because 4.1 is better and faster.
0:11:02 Like, dude, OpenAI sucks at naming these things.
0:11:03 Yeah.
0:11:07 And they had said that they were going to solve that by doing GPT-5, which would be, like, the all-in-one,
0:11:12 the one, you know, the one model to rule them all kind of thing where, like, you just talk to the one model,
0:11:15 and then behind the scenes, it’s routing you to whatever is the best model probably is what they’re doing.
0:11:16 But they didn’t do it.
0:11:17 They, like, delayed it.
0:11:20 They’re like, no, sorry, we’re going to have naming problems for a few more months.
0:11:23 Like, give us, like, two or three more months, and then, you know, that’ll be solved.
0:11:27 So I kind of wonder, like, is 4.5, was it a failed run, or are they just finding that it’s so expensive
0:11:33 that people don’t know how to use it so it makes more sense to improve upon it with some kind of new reasoning model
0:11:37 or with the GPT-5 to have it part of that kind of stack of things that they might recommend?
0:11:41 Like, if you’re trying to write something, maybe behind the scenes you’re using 4.5.
0:11:45 But otherwise, it’s so costly to use it, why would you use it for the thing?
0:11:46 I don’t know.
0:11:48 I mean, 4.5 has been my favorite model.
0:11:53 When I’ve been using ChatGPT lately, I’ve been using 4.5 more than any other model.
0:11:53 Yeah.
0:11:56 But if I had to guess, I think it’s the cost, right?
0:12:01 Because the GPT-4.5, the cost of using it in the API was just, like, mind-blowing.
0:12:04 Like, nobody was using it because it was just way too expensive.
0:12:05 Yeah.
0:12:07 So I think cost was probably the biggest factor.
0:12:12 And supposedly, they rolled out 4.1 inside of the API.
0:12:14 So this is all super confusing, right?
0:12:21 Because 4.5 is available in ChatGPT, but 4.1 is not available in ChatGPT, but it is available in the API.
0:12:24 And they were pushing this, like, good for coding.
0:12:25 Yes.
0:12:27 I think they were pushing this one as the best one for coding.
0:12:33 Yeah, and so it’s kind of confusing, but, like, then if 0.3 is out, I assume it’s got to be better at coding.
0:12:35 So it’s, like, what’s the use case?
0:12:36 Or it’s confusing.
0:12:37 Yeah.
0:12:39 This one is available in the API.
0:12:41 This is GPT-4.1 in the API.
0:12:46 They released 4.1, 4.1 mini, and 4.1 nano.
0:12:49 They outperformed 4.0, 4.0 mini.
0:12:51 Let’s see.
0:12:55 Coding, it scores 54.6% on SWE bench verified.
0:13:00 And it’s an improvement over 4.0 and over 4.5.
0:13:05 So apparently it’s a lot better at coding than 4.5, but 4.5 wasn’t good at coding in the first place.
0:13:06 Right.
0:13:09 It felt more, like, human-like when you were chatting with it, I guess.
0:13:10 Yeah.
0:13:12 But it wasn’t really a model that was good with, like, logic and coding.
0:13:17 One thing I found interesting, though, is, like, yeah, 4.5 has been the best for writing, as far as I can tell.
0:13:21 It has the most personality, but they’ve just kept updating 4.0 behind the scenes, it feels like.
0:13:23 It felt like it just kept getting better and better.
0:13:29 It’s not as good as 4.5 in some ways, but ever since they released even, like, the image thing, it feels like they updated the model somehow.
0:13:30 Like, there’s something.
0:13:31 It’s smarter.
0:13:33 I’m not entirely sure.
0:13:35 You know, they haven’t really talked about that.
0:13:36 But, like, it just appears to be.
0:13:37 So I’m not sure what the takeaway is.
0:13:38 This is all super confusing.
0:13:40 It’s opening eyes fault.
0:13:44 I do have a little bit of a breakdown that we’ll get to once we’ve sort of covered all the announcements.
0:13:44 Yeah.
0:13:47 It sort of tries to organize it a little bit better.
0:13:56 But they did put out this graph, which is intelligence based on multilingual MMLU, massive multitask language understanding.
0:14:00 So this one’s more of, like, a general benchmark of just, like, intelligence.
0:14:06 So we can see that GPT 4.1 nano, the latency, that’s the speed, right?
0:14:09 So, like, how fast from the time you give it the prompt to get your response.
0:14:14 So GPT 4.1 nano is super fast, but not the smartest.
0:14:20 Mini is pretty smart, smarter and faster than 4.0 Mini.
0:14:27 And the new full GPT 4.1 is about the same speed as 4.0, but slightly smarter.
0:14:29 That’s what I’m making from that chart.
0:14:30 Oh, my God.
0:14:31 That’s super confusing.
0:14:34 So this is why they need the GPT 5.
0:14:35 It’s interesting.
0:14:38 They must have really realized, like, you do need different models for different use cases.
0:14:39 Yeah.
0:14:43 And so they’re just like, well, for now, let’s just keep giving them different crazy names.
0:14:45 But then we’ll have to make some kind of smart router.
0:14:47 So hopefully that comes soon.
0:14:47 Yeah.
0:14:55 Well, and again, I think, like, their ultimate goal is you enter a prompt and then it chooses for you the best model to use, right?
0:14:55 Yeah.
0:14:56 Well, that’s what GPT 5.
0:14:57 They’ve already said that’s what it’s going to be.
0:14:57 Okay.
0:14:57 Yeah, yeah.
0:14:59 That’s kind of the goal that they’re trying to get to.
0:14:59 Yeah.
0:15:04 So, you know, so Sweebench, this is basically how good it is at coding.
0:15:13 I believe this is the one where it goes on GitHub and tries to find, like, bugs that need to be fixed and how good is it at fixing them or something like that.
0:15:14 Yeah.
0:15:18 Given a code repository, an issue description must generate a patch to solve the issue.
0:15:22 So GPT 4.1 is way up here.
0:15:29 But one thing I’ve noticed is GPT and OpenAI has completely stopped comparing themselves to, like, external models.
0:15:32 Like, you notice how Gemini 2.5 is not on here.
0:15:35 Notice how Claude 3.7 is not on here.
0:15:40 Like, we don’t know how this compares based on this graph to models that aren’t OpenAI’s models.
0:15:40 Right.
0:15:41 Yeah.
0:15:42 And who knows with coding, too.
0:15:44 Like, things have changed a lot in the last month.
0:15:46 Like, all of a sudden, like, Google took over coding.
0:15:49 Like, Claude was, like, in a huge lead forever.
0:15:58 Then they released 3.7, which seemed good, but then kind of got overambitious because they would try to do tons of things for you and, like, seem to be making more mistakes than 3.5 did.
0:16:04 And then Google came out with, like, 2.5 Pro, their Gemini model, and that’s just been amazing.
0:16:05 Like, people haven’t talked about that enough.
0:16:08 Like, Google has taken over coding with AI in the last month.
0:16:12 I think most people probably don’t realize that, but, like, that has happened in the last month.
0:16:13 Yeah, no, that’s true.
0:16:15 And, I mean, I still do quite a bit of coding.
0:16:19 I’ve kind of been using WindSurf a little more than anything, a little more than Cursor, really.
0:16:25 And, yeah, I’ve been pretty much using Gemini 2.5 the most.
0:16:29 And every once in a while I’ll get hung up and I’ll go to Claude and be like, well, can Claude solve it?
0:16:32 And sometimes Claude can solve random stuff that Gemini can’t.
0:16:32 Yeah.
0:16:42 But I would say, like, 95% of my coding now is done with Gemini 2.5, which will be interesting because now OpenAI just released 03 and 04 Mini,
0:16:53 which this new 03 model is their most powerful model, but they also have 04, which is a bigger number than 03, but not as powerful as 03.
0:16:54 Right.
0:16:57 And, you know, Sam Allman had said a few things about these models.
0:17:04 Well, more recently, he said, the benchmarks you saw for 03, we found some ways to improve upon the model more than we anticipated.
0:17:11 So it’s actually better than what we showed, but also in the past, he hinted at 03 Pro being amazing.
0:17:15 And like you said, off camera, maybe 04 is 03 Pro.
0:17:15 Maybe.
0:17:15 Yeah.
0:17:16 I’m not sure.
0:17:16 Yeah.
0:17:16 Yeah.
0:17:20 Or maybe 03 Pro might be what they launch as GPT-5.
0:17:21 Who knows?
0:17:22 Yeah.
0:17:27 Or 03 Pro could be like in a week or two and it’s only for the professional, you know, users or something like that.
0:17:28 That makes sense.
0:17:28 Yeah.
0:17:32 This 03 is their current most powerful model that we have access to.
0:17:35 It’s available to both Pro and Plus users.
0:17:41 04 Mini, I believe, is a faster, less expensive model, but still a very impressive model.
0:17:42 Yeah.
0:17:44 We can kind of look at some of the charts that they show here.
0:17:50 So they actually had another chart in their video that they’re not showing here for some reason.
0:18:03 But you can see that like they’re 04 Mini with no tools scored 92.7 on competition math and they’re 03 with no tools scored 88.9.
0:18:08 So when it comes to math, I guess 04 Mini is better than 03.
0:18:09 Yeah.
0:18:21 But then they show this other chart in their launch video, which showed 03 with Python and it got like a 99.5% or something crazy like that on the competition math.
0:18:23 Basically like acing it.
0:18:23 Yeah.
0:18:25 That’s why they had to have something like GPT-5.
0:18:31 Like they’re going to have to have like one smart router to like, because all these models are good at different things.
0:18:34 And some of it, you discover it over time.
0:18:37 Like some of it, I’m just like, you learn that like, oh, this model is just slightly better for this one thing.
0:18:39 Yeah.
0:18:39 Yeah.
0:18:47 So we can see here that like 03 and 04 Mini, like when it comes to competition code here are like pretty equal.
0:18:59 And then down here, the PhD level science questions, 03 is actually better at that than 04 Mini, you know, but 04 Mini is better at math, but not as good at science.
0:19:02 How big is the jump in general from like 01 to 03?
0:19:06 So we’ve got the estimated inference cost on this axis here.
0:19:25 And then this is the math benchmark here and we can see 03 Mini, you know, is the lowest cost, but also didn’t score as well where you’ve got 04 Mini high on math here, pretty expensive comparatively, but also, you know, really smart.
0:19:25 Right.
0:19:36 And then we can sort of see it in comparison here to 03, which is not as good, but it’s not showing like the jump from like a one pro to 04, which is kind of frustrating.
0:19:37 Yeah.
0:19:41 Well, I mean, yeah, I think, I think 03 Pro be the more comparable there for a one pro.
0:19:43 I think they’ll probably come out of that soon.
0:19:49 But then here we can see when it comes to GPQA, what does that test for?
0:19:52 Graduate level Google proof.
0:19:58 So basically graduate level questions that you can’t just Google an answer for is essentially what that means.
0:20:08 So on that benchmark, 03 Mini actually performs about the same as 04 Mini high, which is interesting.
0:20:11 They both perform the same.
0:20:13 Wait, so the GPQA is actually physics, by the way.
0:20:15 So it’s like physics questions.
0:20:17 Because I was like, yeah, hard questions.
0:20:20 So cost performance, 01 and 03.
0:20:21 Okay, so here we go.
0:20:22 We can see this is math.
0:20:25 A-I-M-E is math questions, right?
0:20:29 So we can see on math, 03 for inference costs is down here.
0:20:32 03 high is way up here.
0:20:38 And the inference cost is like way cheaper than 01 Pro, which I believe is the same as 01 high.
0:20:39 Yeah.
0:20:41 I don’t know why they’re using different terminology.
0:20:43 I think 01 high is 01 pro.
0:20:58 I mean, I think, you know, remember when we talked about when like reasoning models first came out, I think when we were out in Boston, you know, for the HubSpot inbound conference, you know, we talked about like how things were going to speed up now that there was this new paradigm where it wasn’t just throwing more data at it, but you also could throw more compute power to.
0:21:01 And so we probably would see things improve dramatically faster.
0:21:09 It feels like we’re now seeing that because instead of the model getting, you know, twice as good in like, you know, two years, it’s like now it’s like what, like in three months or something.
0:21:10 Yeah.
0:21:10 Oh yeah.
0:21:15 The, the amount of time between like each of these new releases is just like compressing.
0:21:16 This is so fast.
0:21:17 I mean, here’s one that’s really impressive.
0:21:21 That same GPQ a here, here’s the curve of Oh three.
0:21:24 Look at where Oh one pro is basically down here.
0:21:34 The cost is like the highest possible cost and it’s dumber than Oh three low model or about on par with the Oh three low model here.
0:21:42 And the Oh three high is still maybe slightly cheaper than Oh one medium, but like, look at the jump when in intelligence way up here.
0:21:43 Wow.
0:21:43 Pretty wild.
0:21:47 The speed at which these leaps are happening is pretty crazy.
0:21:48 Yeah.
0:21:50 When we first started the show, we were talking about like how big of a change.
0:21:55 It just wasn’t, I think people probably thought that was hype and it’s actually happening now.
0:21:59 And I think people still don’t understand it’s actually happening and like how big of a deal it is.
0:22:02 Like Sam Altman was interviewed at Ted the other day.
0:22:02 Oh yeah.
0:22:03 I saw that.
0:22:03 Yeah.
0:22:09 And he was talking about how, I think he said that they have 500 million weekly active users now.
0:22:17 And then the Ted guy said, backstage, you told me that your numbers doubled in the last month since the whole Ghibli thing came out.
0:22:22 He was like, that was supposed to be off record, but yeah, you know, yeah, it’s, it’s going well.
0:22:28 I mean, so possibly open the eye is what in the ballpark of maybe 700 plus million users.
0:22:31 I mean, that is just Silicon Valley has never seen this before.
0:22:38 And it’s so wild that Sam Altman used to run YC, like the big incubator for Silicon Valley, that he ends up producing the biggest startup of all time, possibly.
0:22:40 I just wonder why he would want that to be off record.
0:22:43 That’s something that would probably attract investors pretty quickly.
0:22:44 Didn’t know those numbers.
0:22:45 Yeah.
0:22:51 But I think he is in a game with Google and Grok and Anthropic and everyone of like, you want to hype things up.
0:22:54 But also you have to keep some cards secret, right?
0:22:57 Like you don’t want them to know exactly everything.
0:23:00 I mean, there’s probably some misdirection and all kinds of stuff going on.
0:23:00 Yeah.
0:23:00 Yeah.
0:23:02 It’s wild how fast it’s improving.
0:23:09 And one thing that I found interesting is in the interview, he alluded to this being possibly like the most pivotal year for AI.
0:23:13 Like basically saying, and he’s also kind of like saying the best is yet to come kind of thing.
0:23:16 And also talking about the next big thing is scientific discoveries.
0:23:16 Yeah.
0:23:21 And the biggest talk on X over the last week that I’ve seen a lot of the rumors that were going around in some of them
0:23:26 were even shared, like there was like positive things being said by scientists too, kind of alluding to this being possibly true,
0:23:34 is that they’ve had early access to the new open AI models and possibly some new discoveries have been made.
0:23:39 Or at least experiments that they were not able to come up with that are now helping them possibly discover new things in science.
0:23:40 Yeah.
0:23:45 So possibly this is like having like really, really material impact on science for the first time ever.
0:23:46 Yeah.
0:23:46 No.
0:23:48 In fact, let me pull up an article.
0:23:50 I was actually reading this earlier.
0:23:52 It’s on the information here.
0:23:56 Open AI’s latest breakthrough AI that comes up with new ideas.
0:23:56 Yeah.
0:23:59 Now AI is getting good at brainstorming.
0:24:05 This came out a few days ago before this new, you know, O3 and O4 mini came out.
0:24:09 If the upcoming models dubbed O3 and O4 mini perform the way their early testers say they do,
0:24:15 the technology might soon come up with novel ideas for AI customers on how to tackle problems such as designing
0:24:20 or discovering new types of materials or drugs that could attract fortune 500 customers,
0:24:23 such as oil and gas companies and commercial drug developers.
0:24:28 in addition to research lab scientists, the apparent improvements highlight the benefit of AI models
0:24:32 focused on reasoning, which the chat GPT maker debuted in September.
0:24:37 But yeah, basically saying that these new models, the people that got early access have been saying
0:24:43 like this is helping us design and discover new materials and drugs, which is that’s kind of the
0:24:44 holy grail of what people want AI to do.
0:24:45 Yes.
0:24:49 I mean, it looks like we’re already there and we’ve just like scratched the surface of what this new
0:24:51 paradigm with reasoning models can do.
0:24:55 But it seems like the next several years, it’s going to keep accelerating and possibly,
0:24:57 you know, revolutionizing science.
0:24:59 So it’s like just, you know, it’s exciting.
0:24:59 Yeah.
0:25:00 Yeah.
0:25:03 And one of the things about these new models that we haven’t even touched on yet is that
0:25:05 they’re all being built multimodal, right?
0:25:09 So theoretically, you can give them videos, you can give them audio, you can give them text,
0:25:13 you can give them all the modalities, images, anything.
0:25:17 And it will actually be able to understand what’s going on within those.
0:25:21 But also they all have search now and they’re all getting the ability to use tools.
0:25:24 And so this chart I found pretty interesting.
0:25:27 This is one that they showed in their live stream, but they didn’t actually put on their
0:25:27 website.
0:25:33 You can see that when they gave Oh, for many tool use and it was using Python to help with
0:25:38 the math, it scored 99.5 on the competition math.
0:25:43 So like, yeah, basically acing competition math, when you give it access to be able to
0:25:48 use tools, it’s getting to that point where like these benchmarks aren’t good enough anymore.
0:25:53 We need new benchmarks for like, can AI find new discoveries for us?
0:25:54 Like, how do we make those kinds of benchmarks?
0:25:56 Right.
0:25:58 We’re getting more and more to a situation too.
0:26:01 Like when the Oh one came out, it took people a while to realize how good it was or like what
0:26:02 it meant.
0:26:03 You know, Sam Altman said this.
0:26:08 He said, his son is never going to grow up in a time where he’s smarter than AI.
0:26:12 Like for a child who’s being born right now, who’s young, that’s just not going to be possible
0:26:13 in their lifetime.
0:26:14 So like, that’s where we’re at.
0:26:18 Like these models, not in every single way, not in an entirely general way, every single
0:26:23 possible way, but in a lot of ways, they’re already more intelligent than most people.
0:26:25 And that’s just going to continue and accelerate.
0:26:29 So I’m constantly thinking about like, what does this mean for me personally in my business?
0:26:31 And like, what am I going to, how am I going to use it?
0:26:33 I saw Greg Eisenberg tweeting about this the other day.
0:26:36 If he’s telling the real story, if he was just trying to like set up something for social
0:26:40 media, but he’s talking about how to be kind of in a rut because things are changing so fast.
0:26:42 And I feel like there’s so many different things you could do.
0:26:46 And it’s like, what do you do in an environment where things are so dynamically changing?
0:26:48 We’ve never experienced this in human history.
0:26:53 It’s not clear how anyone exactly takes advantage of this because maybe by the time you do something,
0:26:54 it’s already changed again.
0:26:54 Yeah.
0:26:55 Yeah.
0:26:58 We’re entering completely new territory that nobody’s been in yet.
0:27:04 And even all the people that are trying to predict when things are going to happen or what’s going to happen next are all just making predictions.
0:27:07 Like they don’t even necessarily know what’s coming or where this is going to lead.
0:27:08 Yeah.
0:27:12 I’ve been preaching this for a while, but like, I really feel like major corporations are
0:27:13 not taking this seriously enough.
0:27:18 Like, you know, being a corporation who has too many meetings and overly like moves very
0:27:18 slow.
0:27:20 That’s a bad spot to be in.
0:27:24 I think people don’t realize how smart Elon Musk has been in this area where he’s like
0:27:26 really like made things more lean and like move fast.
0:27:31 Every big company should be doing that because otherwise, how do you survive in an environment
0:27:33 where things are dramatically changing?
0:27:33 Yeah.
0:27:33 Yeah.
0:27:37 Did you see Toby Lucky’s statement from Shopify?
0:27:38 Yeah.
0:27:38 Yeah.
0:27:42 I mean, he’s basically saying what you said, but the big point that he made is if you guys
0:27:48 want to hire anybody else in Shopify, you must prove that what you’re hiring for can’t be
0:27:49 done with AI first.
0:27:50 Right.
0:27:54 If you can prove AI cannot do it, then okay, you can make the hire, but you must prove to
0:27:56 me that AI can’t do it first.
0:27:56 Right.
0:28:00 And it’s also super clever because he’s also kind of like baking in AI into the culture
0:28:03 of the company where it’s like, look, if you’re involved in the company, you have to be thinking
0:28:06 about how you could be using AI to make the company better.
0:28:06 Yeah.
0:28:10 Because otherwise, how could you do that test if you’re not like actually using AI and like
0:28:11 actively understanding what it can do?
0:28:13 So that’s super brilliant.
0:28:16 And I don’t think he’s coming from a place of like, AI is going to replace all of you.
0:28:20 I think he’s coming from a place of like, this company can be so much more efficient
0:28:24 and so much better and get things done so much more smoothly if we’re all leveraging
0:28:25 AI together.
0:28:28 He’s talking about like an AI first company, like even though, even though they’re a large
0:28:32 org restructuring already to be an AI first company.
0:28:34 And most companies are not realizing this.
0:28:38 I mean, I, I think almost all major companies should be reorganizing right now ahead of time,
0:28:42 but you know, like people always, you know, they’re more reactive and proactive, like especially
0:28:46 big organizations, like all the companies, they will realize this like five or 10 years from
0:28:47 now.
0:28:50 And then some of them will not survive because of that, but it’s gonna be a wild time.
0:28:51 Yeah.
0:28:53 So I want to bring this one home with this.
0:28:54 Yeah.
0:28:59 So I actually asked open AI is deep research to sort of break down all of the models,
0:29:04 like help get past the confusion, do the deep research on which models are good at what,
0:29:09 and then give me like a ranked order of basically which was the dumbest to the smartest of all of
0:29:11 the available models right now.
0:29:21 It actually took, let’s see, 16 minutes for even a chat GPT to figure out, um, its own dang models
0:29:25 with how confusing it all is, but here’s the breakdown.
0:29:32 So GPT 3.5 is great at basic conversations, but substantially lower reasoning and knowledge
0:29:32 than the others.
0:29:38 Then GPT four, which came out in March, 2023 was a big leap from 3.5.
0:29:44 And then you have GPT four turbo slash GPT four Oh, which apparently are sort of tied.
0:29:51 Then you have GPT 4.1 nano, which is one that came out this week, which is a very small, but
0:29:57 surprisingly capable model, 80% of GPT fours ability, basically.
0:30:02 Then you have a GPT 4.1 mini slash GPT four Oh.
0:30:04 So GPT four one mini was one that came out this week.
0:30:08 GPT four Oh is the one that’s sort of been the default model for a while.
0:30:15 inside of chat GPT, but GPT four one mini, it says matches or exceeds GPT four Oh’s intelligence.
0:30:17 So both are kind of tied.
0:30:24 Then you have GPT 4.5, which was the one that they codenamed Orion GPT 4.5.
0:30:26 I know it’s been available in the $200 a month plan.
0:30:30 I can’t remember if it was available in the $20 a month plan, but that one’s been one that
0:30:32 was sort of more human.
0:30:37 like, I guess it’s good at writing stories, good at when you just want to have like a
0:30:37 chat with an AI.
0:30:42 That was the one that I always felt like was the best at having like a very human like conversation.
0:30:44 Then you have open AI’s Oh one.
0:30:49 This was their first reasoning model where it actually sort of thought through after you gave it the prompt.
0:30:53 And so during inference, it sort of did that sort of chain of thought thinking.
0:30:55 This was the first model we saw that was doing that.
0:31:05 Then you have Oh four mini, which was, I believe one that also came out this week is a little bit better than just the standard Oh one.
0:31:09 So it’s a new efficient reasoning model when run in high compute mode.
0:31:15 It’s extremely capable, likely outperforming Oh one and approaching the top models in STEM benchmarks.
0:31:26 Then you have GPT 4.1 full, which is another one that came out this week, which is open AI’s latest GPT model only available in the API right now, not available in chat GPT.
0:31:37 And this one’s more tuned for coding and following instructions, but this is why it’s a little bit confusing because we can see GPT 4.1 is actually a smarter model than GPT 4.5.
0:31:42 So even though the numbering scheme went backwards, it’s actually a more powerful model.
0:31:49 And then you have open AI Oh one pro, which for the longest time was open AI’s top dog model.
0:31:54 If you got into chat GPT and you had the $200 a month plan, Oh one pro was the best of the best.
0:31:58 It just kind of crushed anything you threw at it.
0:32:02 It was really good at coding took forever to get into the API, but that was their best for a long time.
0:32:06 And then you have open AI Oh three, which was also one that was just released this week.
0:32:08 It is available in chat GPT.
0:32:13 It’s available both in pro and plus so 20 and $200 a month plans.
0:32:19 This one is their new best, most powerful model they’re making available right now.
0:32:20 That’s open AI Oh three.
0:32:28 And then this list says their most intelligent, best possible model you can use is Oh three with tool use turned on.
0:32:35 So allowing Oh three to use tools like Python and things like that to help it with its, you know, math and coding and logic and things like that.
0:32:38 When you use Oh three with tools, it’s their best model.
0:32:40 And then also GPT 4.1.
0:32:44 They’ve also sort of tied with it depending on the use case.
0:32:46 So insanely confusing.
0:32:46 Yeah.
0:32:50 But this is the order it ranked it in based on all the deep research.
0:32:50 Yes.
0:32:54 I mean, so thankfully, like people won’t have to worry about this soon.
0:32:57 So like, hopefully in like two months, you won’t have to worry about this GPT five, hopefully be out.
0:32:59 It’ll manage all that for you.
0:33:00 You just talk to one model.
0:33:05 But right now, I think their descriptions actually are pretty good in line with my experience.
0:33:11 Like, I think there’s like three models people really would the average people would use right now, like for Oh, that’s what most people would use.
0:33:13 Like I said, that model keeps getting better.
0:33:13 Yeah.
0:33:16 And if you’re just using ChatGPT free, that’s what you’re using.
0:33:17 You’re using 4.0.
0:33:17 Yeah.
0:33:18 4.0 is great.
0:33:19 It just kept getting better.
0:33:23 They haven’t really like made a big fuss about it, but it seemed to have just kept getting smarter in some way.
0:33:25 I don’t think that’s entirely reflected in the benchmarks.
0:33:28 4.5 still seems to be the best for writing.
0:33:31 So if you’re wanting to edit something or write something, 4.5 is the best.
0:33:33 That one’s also getting phased out.
0:33:34 So use it while you can.
0:33:36 Use it while you can.
0:33:38 I think it’ll still be there in five.
0:33:39 I had to guess behind the scenes.
0:33:48 And then O3, which I haven’t used yet, based on the benchmarks and my understanding how the models work, I would assume that’s going to be the best if you have any complicated questions.
0:33:52 Like if there’s something that you’re thinking like, oh, the model might give me too surface level of an answer.
0:33:54 It’s not taking the time to think it through.
0:34:00 If you’re asking something complex, you probably want to hand it to O3 and give it the time to really think through its response to you.
0:34:03 And that’s also probably the one I would imagine you’re going to go to for coding as well.
0:34:04 Yeah, probably so.
0:34:06 You know, I haven’t had a chance to test any of it with coding yet.
0:34:12 There’s one last thing that we didn’t actually bring up about the new models that they just released, and that’s the context window.
0:34:12 Yeah.
0:34:15 One of these models is actually a 1 million token context window.
0:34:17 Is it a real one?
0:34:20 Like I remember that was a part of the whole fiasco with Meta as well.
0:34:21 It was like they said like 2 million context window.
0:34:27 And then apparently at all the benchmarks, it would like start to like fall apart after like 128K context or something.
0:34:28 I don’t know.
0:34:33 I actually kind of trust OpenAI’s word on this a little bit more than I trust Meta’s word on it.
0:34:33 Yeah, me too.
0:34:34 Me too.
0:34:35 Because that was just ridiculous.
0:34:36 Like people tested it.
0:34:40 It’s like they were hyping up like 2 million contacts or 4 million or whatever it was they hyped up.
0:34:43 And then the people tested it.
0:34:47 And as soon as it went beyond like 128K contacts, like the model just like became really dumb.
0:34:49 Yeah, I heard the same thing.
0:34:55 I mean, I’m sure there’ll be some diminishing returns like O1 Pro even like, you know, that’s why I was using a repo prompt is like, yeah, it could take a lot of context.
0:35:02 But you do want to kind of be careful about what you give it because if you give it more exactly what it needs, you do tend to get better results back.
0:35:09 But yeah, these context windows are getting to a point where like context length is not going to be an issue for very much longer.
0:35:09 Yeah.
0:35:13 Like you’re going to be able to plug in entire code bases and it’s going to be able to read the whole dang thing.
0:35:20 Yeah, and it is interesting that like Sam Altman kind of admittedly, well, he’s at number one, like people were asked about like DeepSeek and he’s like, well, it hasn’t impacted our growth at all.
0:35:22 So not really too worried about it.
0:35:28 But also he was saying that like moving forward, it’s probably more about the things they build on top of the models.
0:35:31 He thinks they’re going to continue to be in the lead.
0:35:35 He said, but that lead is going to narrow and it’s going to be like a small lead in the future.
0:35:37 So it’s all about what they build on top of it.
0:35:41 You know, it did seem early on that like Open the Eye was going to be incredibly ahead of the game.
0:35:44 But now it feels like there’s going to be lots of people who are pretty close to them.
0:35:45 So here we go.
0:35:50 So as of April 2025, here’s the context window sizes for OpenAI’s latest models.
0:35:55 So it was the GPT 4.1 series, the one that’s only available in the API.
0:35:56 Yeah.
0:36:03 That one is a 1 million token context window, which includes GPT 4.1, 4.1 Mini and 4.1 Nano.
0:36:05 So those are a million tokens.
0:36:08 GPT 4.0 is still 128,000.
0:36:11 O3 Mini supports up to 200,000.
0:36:16 O3 Full, likely maintaining or exceeding 200,000.
0:36:18 But it doesn’t actually say here.
0:36:22 O4 Mini released along O3 is a reduced version of a successor.
0:36:27 Specific context window sizes for O4 Mini have not been detailed.
0:36:31 So yeah, the O3 and O4, they haven’t actually said what the context window was.
0:36:38 But the new 4.1, which is the one that’s going to replace 4.5, is the million token context window.
0:36:42 So yeah, that’s pretty much a breakdown of all the OpenAI news that came out over the last few days.
0:36:46 There was one smaller update that’s just like a quality of life thing.
0:36:53 If you are generating images inside of OpenAI, they created this library page where you can see all the images that you’ve generated in just like one spot.
0:36:56 You know, kind of cool quality of life thing that they’ve added as well.
0:36:59 Yeah, there’s another move towards killing Mid-Journey possibly.
0:37:00 Yeah, very true.
0:37:02 Yeah, you know, Mid-Journey 7 came out.
0:37:06 Like people were expecting it was going to be way better at understanding images.
0:37:08 And it’s like slightly better.
0:37:08 Yeah, yeah.
0:37:10 It is still the most beautiful model.
0:37:11 And it got slightly more beautiful.
0:37:12 Yeah.
0:37:15 It more consistently beautiful and, you know, and things like that.
0:37:20 But it seems like they did not nail the thing where, oh, the characters are going to be completely consistent.
0:37:22 Like you can like reuse the characters.
0:37:22 Yeah.
0:37:24 As far as I can tell, they did not nail that.
0:37:26 Yeah, I played around with V7 a little bit.
0:37:29 I was actually about to make a video on it and didn’t even release the video.
0:37:32 Like I only want to put out videos about stuff that I’m excited about.
0:37:36 And like I was playing with it while recording and like it wasn’t really exciting me.
0:37:38 I’m like, it doesn’t feel like that big of a leap to me.
0:37:43 Yeah, the only cool part of it was the turbo thing where you can like rapidly try ideas.
0:37:44 That’s cool.
0:37:44 Yeah.
0:37:45 And you can do voice mode.
0:37:51 And so that was actually the only cool thing about version 7 was you could just press a voice mode button and just start chatting.
0:37:54 And as you’re chatting, it starts creating the images super fast.
0:37:55 That was cool.
0:37:57 Yeah, I actually didn’t play with that mode.
0:37:58 So that’s one I’ll have to mess with a little bit.
0:37:59 That’s the actually cool part of it.
0:38:00 Like everything else is like whatever.
0:38:01 Yeah.
0:38:03 I didn’t find it that impressive.
0:38:03 Yeah.
0:38:05 Ideogram has gotten really, really good.
0:38:08 The various Leonardo models have gotten really, really good.
0:38:08 Yeah.
0:38:11 Obviously, OpenAI’s model is really, really good.
0:38:18 Like all of these models have sort of caught up to each other where like mid-journey is just sort of in the mix with them as opposed to being the leader, you know?
0:38:19 Yeah.
0:38:29 But like I said in the past, I think people don’t understand like with OpenAI’s models, you know, with the reasoning models, they’re going to be able to actually understand what’s going on in the images and the videos they create.
0:38:29 Right.
0:38:32 Like that’s going to be the unlock, just like making something beautiful.
0:38:35 I’m sure you can do that, but do you understand what the hell is in the thing you just created?
0:38:36 And can you modify that?
0:38:38 That’s the thing that’s interesting.
0:38:42 And I think, I think OpenAI is like the one who’s like by far the leader in that right now.
0:38:43 Yeah.
0:38:43 Yeah.
0:38:44 They’re on the forefront of it for sure.
0:38:45 Yeah.
0:38:52 But yeah, I mean, I think that’s a pretty good breakdown of everything that’s been happening in the world of OpenAI this week.
0:38:52 Yeah.
0:38:58 We went off on a few little tangents, but, you know, hopefully we brought it all home by sort of showing you the ranking of intelligence.
0:39:02 And I think Nathan, you did a good job of saying like, all right, these are the three models you’re probably going to use.
0:39:14 So, you know, hopefully you found this informative and helpful trying to declutter your mind on all of these different OpenAI models because it’s still confusing to us.
0:39:17 I mean, we’re paying attention to it every day and we still get confused.
0:39:19 So, you know, hopefully this helped a little bit.
0:39:23 And thank you so much for tuning in to this episode.
0:39:24 Hopefully you enjoyed it.
0:39:27 If you like it, make sure that you subscribe ideally on YouTube.
0:39:30 That’s where we’re really trying to primarily grow the show.
0:39:35 But if you prefer audio, we’re available on Spotify and iTunes and wherever you listen to podcasts.
0:39:38 Thank you once again for tuning in and hopefully we’ll see you in the next one.
0:39:39 Goodbye.
0:40:01 Bye.
0:00:04 I’m Matt Wolfe.
0:00:06 I’m here with Nathan Lenz.
0:00:12 And today we’re going to try to do our best to demystify all of these recent launches with
0:00:19 OpenAI because they’ve released 4.5 and then 4.1 and then 04 and then 03.
0:00:21 And it’s just getting really, really confusing.
0:00:27 So this episode is designed to break it all down for you, tell you which each of these
0:00:32 models are good for, what they’re not good for, and we’ll even rank them from dumbest
0:00:35 to smartest at the end and tell you which ones to use for which things.
0:00:40 So without further ado, let’s just go ahead and jump in and talk about all these recent
0:00:40 OpenAI announcements.
0:00:47 I think probably the best place to start is with this one right here, because this is probably
0:00:52 the first piece of news in this sort of sprint that OpenAI is on of releasing news, which
0:00:56 is that they just rolled out this memory feature in OpenAI.
0:01:02 And to me, this is probably one of the coolest, most valuable features they’ve rolled out.
0:01:04 I found it really useful.
0:01:04 Yeah.
0:01:08 I know Nathan, before we hit record, you were saying something to the same effect of like,
0:01:10 out of everything, the memory is probably the coolest.
0:01:11 Yeah.
0:01:15 So it’s funny, the team at Grok, they reached out to me about what should they build, like
0:01:16 maybe like two months ago.
0:01:18 And I was like, focus on memory.
0:01:23 Nailed that first, because OpenAI has like a half-baked version of memory.
0:01:26 They’ll probably eventually get there, but who knows if it’s a priority or not.
0:01:28 If it’s not, you should get there first.
0:01:29 Right, right.
0:01:33 Unfortunately for Elon Musk and Grok, OpenAI definitely got there first.
0:01:35 And I think it’s going to be one of the biggest things long term.
0:01:38 It’s the kind of thing where people don’t realize why it’s important now.
0:01:40 But over time, it’s going to be clear why it’s important.
0:01:44 You know, I use ChatGPT a lot, like for so many different things, like personal stuff,
0:01:45 business, everything.
0:01:48 I use it for so many different things, probably more than the average person.
0:01:50 And I asked it to like, explain what do you know about me?
0:01:52 And I’m not going to show you what it said.
0:01:54 Something that was so revealing.
0:01:55 Was it not nice?
0:01:56 No, it was nice.
0:01:57 It was nice.
0:01:59 You know, because it knows a lot about my life too.
0:02:00 My life is pretty crazy.
0:02:05 And just like some of the insights that drew out of that were just kind of shocking and inspiring.
0:02:08 Honestly, it was like, yeah, I have done a lot of great things.
0:02:10 I have went through a lot of hard times in my life and all this stuff.
0:02:14 And it was just, it seemed to know things about what I want to accomplish in my life.
0:02:18 Things that I don’t think I’ve told it that were just kind of like, that’s me.
0:02:24 And so the idea that AIs are already starting to actually understand me maybe better than I understand myself.
0:02:24 Yeah.
0:02:26 And I mean, you’ve probably had a lot of chats.
0:02:33 I know you’ve mentioned that you’ve done even like voice chats where you go for like walks and just sort of like unload your thoughts into the voice chat.
0:02:35 And I mean, all of that is in memory now.
0:02:38 I mean, and just to be clear, you can turn off the memory.
0:02:41 If this feature freaks you out, there is an option to turn it off.
0:02:49 So like you don’t have to let it remember everything, but like, you know, all of those voice conversations where you’re just sort of like brain dumping.
0:02:53 That’s all in its context now when you chat with it.
0:02:54 Yeah.
0:02:57 I think long-term that’s how AI is going to get really useful for people.
0:03:04 Cause like people think of it right now, it’s just chat to you where you go in there and you copy and paste something in, maybe have it help you in some way, uh, or you just chat with it.
0:03:11 But eventually these models are going to understand maybe what you want or what you want to accomplish better than you yourself understand that.
0:03:13 Or even before you realize you want it.
0:03:14 Or before you do.
0:03:14 Right.
0:03:16 And this is like, this used to be like the dream.
0:03:19 Like when I first got involved in Silicon Valley, like we were all dreaming of this stuff.
0:03:21 We’re like, how do we make the web better?
0:03:23 Like, well, you got to make it more personalized while we actually understand people.
0:03:25 It’s like, how do you do that?
0:03:27 I guess machine learning or something.
0:03:28 And like, we didn’t know how to actually do it.
0:03:33 And people started doing it with like a very basic version of like, oh, well, Amazon’s kind of pulled off something like that.
0:03:36 Cause they kind of get what you want, but it’s not perfect.
0:03:37 Same as Netflix, not perfect.
0:03:42 Eventually all of this stuff, it’s just going to know about the kind of stuff you want to accomplish or do in your life.
0:03:46 And I think that’s where we’re going to get to a point where AI is going to be really helpful.
0:03:49 Anything you’re wanting to do, maybe it’s going to recommend it to you.
0:03:59 Beforehand, like you said, it’s going to know you or it’s going to just know the kind of stuff that you want to do and be able to help you better with more personalized help versus just some generic stuff that it says to everyone.
0:03:59 Yeah.
0:04:08 I mean, sort of random side topic, but since you mentioned Netflix, apparently Netflix is also going to be using OpenAI’s AI tech inside there.
0:04:12 So it will actually even better understand you and what movies to recommend.
0:04:13 But that makes sense.
0:04:17 Like I said, when I first got involved in Silicon Valley, everyone was trying to figure out how do you do that?
0:04:23 Cause it felt like the natural evolution of software products was to understand users and what they wanted better on a personal level.
0:04:25 And no one really nailed it.
0:04:33 Like I said, Amazon and Netflix got somewhat close and everyone’s just been like tweaking on what Amazon Netflix has been doing for like 10, 20 years now or something like that.
0:04:34 So yeah, it makes sense.
0:04:34 Yeah.
0:04:41 Well, I actually like the very first thing when this new memory feature rolled out was there was a button that says, do you want to try it?
0:04:42 And I said, yeah, sure.
0:04:42 Let’s try it.
0:04:45 And it automatically pasted in this prompt.
0:04:46 Describe me based on all our chats.
0:04:47 Make it catchy.
0:04:49 This is what it said about me.
0:04:50 I don’t mind sharing mine.
0:04:58 It says you’re Matt Wolf, a future forward creator, a high trailblazer, and the go-to tech Sherpa for over a half a million subscribers.
0:05:00 You’re not just riding the wave of innovation.
0:05:02 You’re helping build the surfboard.
0:05:07 Your YouTube channel is a hub for creators, entrepreneurs, and curious minds hungry for what’s next in AI and tech.
0:05:14 You break down the complex, spotlight the useful, and always bring the human angle all while rocking a retro futuristic vibe.
0:05:21 Whether it’s a short, punchy, real, or deep dive 20-minute banger, your goal is always the same.
0:05:24 Empower your audience to do more, create faster, and stay ahead.
0:05:29 Off camera, you’re a family guy with a white pup named Bowie, a deep love for camping,
0:05:32 and a soft spot for the San Diego Padres.
0:05:35 You’ve got a musician’s soul, a builder’s brain, and a storyteller’s heart.
0:05:39 Even when burnout hits or the algorithm plays rough, you keep showing up.
0:05:41 Driven by purpose, not just metrics.
0:05:43 You’re not just documenting the future.
0:05:45 You are part of shaping it.
0:05:48 I was pretty flattered by that, so I don’t mind sharing it.
0:05:49 Yeah, yeah.
0:05:53 I mean, like, parts of that were kind of generic, and then parts of it seemed like it was, you know, kind of personalized view.
0:06:00 I have a feeling I probably talked about more personal stuff with it because the stuff it said to me felt more personal than that.
0:06:00 Yeah.
0:06:06 And I mean, I also even gave it, like, a primer in the custom instructions, like, here’s who I am, and here’s, like, a general outline of my life.
0:06:08 Yeah, yours was probably like, and you keep refusing to get that wart checked out.
0:06:09 Go get it.
0:06:19 So, you know, I probably provided it a lot more about myself, but it said a lot of things, you know, it’s just, like, reading between the lines, but it said a lot of things that I did not tell it, though, like, exactly.
0:06:29 Like, it, like, kind of, like, understood things about me, about how I feel about myself, about my accomplishments, my failures, things I want to accomplish before I die, like, all these kind of things.
0:06:30 It was just kind of shocking.
0:06:32 And it was stuff you didn’t actually tell it?
0:06:34 It sort of figured it out?
0:06:34 Some of it.
0:06:35 Some of it.
0:06:39 And I was like, holy, you know, it’s, like, it’s connecting the dots.
0:06:45 It’s, like, understood my personality somewhat and the things that I get inspired by or that I want to do or that bother me or whatever.
0:06:45 Right.
0:06:52 But it definitely connected some dots and said some things that I did not explicitly tell it, which was just, holy crap, that’s amazing.
0:06:53 Yeah, yeah.
0:06:59 For me, this one feels, like, pretty on par with, like, the various discussions I’ve had, right?
0:07:02 I have, you know, talked about my YouTube channel a lot, right?
0:07:06 I use it a lot for, hey, I need an idea for this YouTube video.
0:07:07 Here’s the transcript.
0:07:08 Help me come up with a title.
0:07:09 All right.
0:07:11 Help me come up with some concepts for thumbnails, right?
0:07:14 So I go to it a lot for, like, ideation around my YouTube channel.
0:07:18 So it obviously knows a lot about the fact that I do stuff on YouTube.
0:07:18 Yeah.
0:07:25 It has all the transcripts of my YouTube videos because I always copy and paste them in and go, hey, how should I title this thing?
0:07:25 Right.
0:07:30 So a lot of this is based on probably transcripts from my YouTube videos.
0:07:30 Right.
0:07:36 It’s kind of random that, like, the one sort of personal thing it picked out about me is that I have a dog named Bowie.
0:07:36 Right.
0:07:40 Not that I’m, like, married with two kids or, you know, things like that.
0:07:41 Right.
0:07:43 But just I have a white dog named Bowie.
0:07:46 I’m like, I don’t actually remember telling you that, but it’s true.
0:07:49 So I’m sure it was in, like, one of my videos or something somewhere.
0:07:49 Right.
0:07:50 Yeah.
0:07:56 I mean, mine was more personal and, like, to the point where I didn’t cry, but it made me feel emotional, like, reading it.
0:07:57 And I was like, that is crazy.
0:07:58 And just imagine where that goes.
0:08:00 Like, we always say that’s the worst it’s ever going to be.
0:08:00 Right.
0:08:01 Right.
0:08:01 Right.
0:08:07 And the more you use it, the more it’s going to know about you and the better response you’re going to get on things like this.
0:08:11 We’re totally going to have the thing that, like, you know, in the future, when you pass away, you’ll have, like, a memory of who you were.
0:08:17 It won’t be perfect, but, God, it’ll probably be, like, a 90% of the essence of, like, who you were as a person.
0:08:21 And that is just shocking that, like, that technology is, like, basically, it’s almost there.
0:08:22 It’s, like, it’s right there.
0:08:24 It’s getting really close.
0:08:27 And, you know, this was sort of just the first announcement.
0:08:29 This announcement came out on April 10th.
0:08:29 Yeah.
0:08:30 Right.
0:08:33 So this was sort of, like, the first in a series of, like, almost every day.
0:08:33 Yeah.
0:08:36 The most important thing Grok could have built, they beat them to it.
0:08:37 Small little announcement.
0:08:39 And then on to the real things.
0:08:39 Yeah.
0:08:39 Yeah.
0:08:42 They didn’t even seem to make much of a big deal about it.
0:08:44 I don’t think, like, they even did a sort of live stream for this one.
0:08:47 Almost everything OpenAI does a live stream about in this one.
0:08:48 They didn’t even.
0:08:50 They’re just like, here, here’s a new feature that we rolled out.
0:08:53 Well, they’re talking about doing a social network and stuff now.
0:08:57 Which, if they do that, memory is a core feature of that of actually understanding who the people are.
0:08:57 Right?
0:08:59 That’s how you would build some kind of new social network.
0:09:03 Because it’d be highly customized based on the personality of the people in it.
0:09:03 Yeah.
0:09:07 I mean, I actually haven’t read too much into the whole social network thing.
0:09:10 I just pulled up an article from Reuters about it.
0:09:14 But, yeah, that’s another announcement that has sort of bubbled up.
0:09:19 Is that OpenAI is working on an X-like social media network, which is interesting, right?
0:09:25 Because there’s, like, a pretty big feud between Sam Altman and Elon Musk.
0:09:25 Yeah.
0:09:26 But it makes sense, though.
0:09:28 This is why I told the Gronk guys, like, that’s why they should do it.
0:09:30 It’s because they already have all that data.
0:09:30 Yeah.
0:09:32 About what people are sharing and how they’re interacting with people.
0:09:36 And so that would be their advantage over OpenAI is that data.
0:09:36 Yeah.
0:09:37 So what makes this one different?
0:09:39 Have you read into this one at all?
0:09:40 About what OpenAI is going to do?
0:09:41 Yeah.
0:09:43 Like, why would people use this over something like X?
0:09:45 It seems like it’s all rumors at this point.
0:09:46 I’m not sure if it’s actually even confirmed.
0:09:47 Yeah.
0:09:50 So it looks like it’s all sort of speculation at the moment.
0:09:50 Yeah.
0:09:51 Speculation.
0:09:53 I would assume that they wouldn’t be going after X.
0:09:55 They would be going after Facebook, if I had to guess.
0:09:59 Because Facebook is that kind of product where it’s all about, like, all your friends,
0:10:01 all these connections that you have, like, real life.
0:10:05 And I feel like OpenAI would be in a great spot to, like, revolutionize that kind of product
0:10:07 versus, like, competing with X.
0:10:08 Yeah.
0:10:08 I’m not sure.
0:10:10 I think there’s not a whole lot of details yet.
0:10:11 Yeah.
0:10:11 Super interesting.
0:10:14 We’ll kind of keep an eye on how that one plays out.
0:10:15 I’m sure Mark is, like, we’re super happy.
0:10:17 Like, his AI effort is totally floundering.
0:10:20 And then now Sam Altman may be coming to kill him.
0:10:22 I think it’s more likely to kill Facebook than X.
0:10:24 So, yeah, that’s a tough spot to be in.
0:10:24 Yeah, yeah.
0:10:29 So one of the other things that OpenAI has been doing is they’ve been releasing these new models,
0:10:31 but they’re also sunsetting a bunch of other models, right?
0:10:37 So apparently GPT-4 is going to be phased out on April 30th.
0:10:44 And then also GPT-4.5, which up until this week was, like, the newest model they released,
0:10:46 is also going to get phased out.
0:10:51 They sent out an email to all of the OpenAI API users, or maybe it was everybody in ChatGPD.
0:10:52 I don’t know.
0:10:59 But they sent out an email basically saying, we’re going to phase out 4.5 because 4.1 is better and faster.
0:11:02 Like, dude, OpenAI sucks at naming these things.
0:11:03 Yeah.
0:11:07 And they had said that they were going to solve that by doing GPT-5, which would be, like, the all-in-one,
0:11:12 the one, you know, the one model to rule them all kind of thing where, like, you just talk to the one model,
0:11:15 and then behind the scenes, it’s routing you to whatever is the best model probably is what they’re doing.
0:11:16 But they didn’t do it.
0:11:17 They, like, delayed it.
0:11:20 They’re like, no, sorry, we’re going to have naming problems for a few more months.
0:11:23 Like, give us, like, two or three more months, and then, you know, that’ll be solved.
0:11:27 So I kind of wonder, like, is 4.5, was it a failed run, or are they just finding that it’s so expensive
0:11:33 that people don’t know how to use it so it makes more sense to improve upon it with some kind of new reasoning model
0:11:37 or with the GPT-5 to have it part of that kind of stack of things that they might recommend?
0:11:41 Like, if you’re trying to write something, maybe behind the scenes you’re using 4.5.
0:11:45 But otherwise, it’s so costly to use it, why would you use it for the thing?
0:11:46 I don’t know.
0:11:48 I mean, 4.5 has been my favorite model.
0:11:53 When I’ve been using ChatGPT lately, I’ve been using 4.5 more than any other model.
0:11:53 Yeah.
0:11:56 But if I had to guess, I think it’s the cost, right?
0:12:01 Because the GPT-4.5, the cost of using it in the API was just, like, mind-blowing.
0:12:04 Like, nobody was using it because it was just way too expensive.
0:12:05 Yeah.
0:12:07 So I think cost was probably the biggest factor.
0:12:12 And supposedly, they rolled out 4.1 inside of the API.
0:12:14 So this is all super confusing, right?
0:12:21 Because 4.5 is available in ChatGPT, but 4.1 is not available in ChatGPT, but it is available in the API.
0:12:24 And they were pushing this, like, good for coding.
0:12:25 Yes.
0:12:27 I think they were pushing this one as the best one for coding.
0:12:33 Yeah, and so it’s kind of confusing, but, like, then if 0.3 is out, I assume it’s got to be better at coding.
0:12:35 So it’s, like, what’s the use case?
0:12:36 Or it’s confusing.
0:12:37 Yeah.
0:12:39 This one is available in the API.
0:12:41 This is GPT-4.1 in the API.
0:12:46 They released 4.1, 4.1 mini, and 4.1 nano.
0:12:49 They outperformed 4.0, 4.0 mini.
0:12:51 Let’s see.
0:12:55 Coding, it scores 54.6% on SWE bench verified.
0:13:00 And it’s an improvement over 4.0 and over 4.5.
0:13:05 So apparently it’s a lot better at coding than 4.5, but 4.5 wasn’t good at coding in the first place.
0:13:06 Right.
0:13:09 It felt more, like, human-like when you were chatting with it, I guess.
0:13:10 Yeah.
0:13:12 But it wasn’t really a model that was good with, like, logic and coding.
0:13:17 One thing I found interesting, though, is, like, yeah, 4.5 has been the best for writing, as far as I can tell.
0:13:21 It has the most personality, but they’ve just kept updating 4.0 behind the scenes, it feels like.
0:13:23 It felt like it just kept getting better and better.
0:13:29 It’s not as good as 4.5 in some ways, but ever since they released even, like, the image thing, it feels like they updated the model somehow.
0:13:30 Like, there’s something.
0:13:31 It’s smarter.
0:13:33 I’m not entirely sure.
0:13:35 You know, they haven’t really talked about that.
0:13:36 But, like, it just appears to be.
0:13:37 So I’m not sure what the takeaway is.
0:13:38 This is all super confusing.
0:13:40 It’s opening eyes fault.
0:13:44 I do have a little bit of a breakdown that we’ll get to once we’ve sort of covered all the announcements.
0:13:44 Yeah.
0:13:47 It sort of tries to organize it a little bit better.
0:13:56 But they did put out this graph, which is intelligence based on multilingual MMLU, massive multitask language understanding.
0:14:00 So this one’s more of, like, a general benchmark of just, like, intelligence.
0:14:06 So we can see that GPT 4.1 nano, the latency, that’s the speed, right?
0:14:09 So, like, how fast from the time you give it the prompt to get your response.
0:14:14 So GPT 4.1 nano is super fast, but not the smartest.
0:14:20 Mini is pretty smart, smarter and faster than 4.0 Mini.
0:14:27 And the new full GPT 4.1 is about the same speed as 4.0, but slightly smarter.
0:14:29 That’s what I’m making from that chart.
0:14:30 Oh, my God.
0:14:31 That’s super confusing.
0:14:34 So this is why they need the GPT 5.
0:14:35 It’s interesting.
0:14:38 They must have really realized, like, you do need different models for different use cases.
0:14:39 Yeah.
0:14:43 And so they’re just like, well, for now, let’s just keep giving them different crazy names.
0:14:45 But then we’ll have to make some kind of smart router.
0:14:47 So hopefully that comes soon.
0:14:47 Yeah.
0:14:55 Well, and again, I think, like, their ultimate goal is you enter a prompt and then it chooses for you the best model to use, right?
0:14:55 Yeah.
0:14:56 Well, that’s what GPT 5.
0:14:57 They’ve already said that’s what it’s going to be.
0:14:57 Okay.
0:14:57 Yeah, yeah.
0:14:59 That’s kind of the goal that they’re trying to get to.
0:14:59 Yeah.
0:15:04 So, you know, so Sweebench, this is basically how good it is at coding.
0:15:13 I believe this is the one where it goes on GitHub and tries to find, like, bugs that need to be fixed and how good is it at fixing them or something like that.
0:15:14 Yeah.
0:15:18 Given a code repository, an issue description must generate a patch to solve the issue.
0:15:22 So GPT 4.1 is way up here.
0:15:29 But one thing I’ve noticed is GPT and OpenAI has completely stopped comparing themselves to, like, external models.
0:15:32 Like, you notice how Gemini 2.5 is not on here.
0:15:35 Notice how Claude 3.7 is not on here.
0:15:40 Like, we don’t know how this compares based on this graph to models that aren’t OpenAI’s models.
0:15:40 Right.
0:15:41 Yeah.
0:15:42 And who knows with coding, too.
0:15:44 Like, things have changed a lot in the last month.
0:15:46 Like, all of a sudden, like, Google took over coding.
0:15:49 Like, Claude was, like, in a huge lead forever.
0:15:58 Then they released 3.7, which seemed good, but then kind of got overambitious because they would try to do tons of things for you and, like, seem to be making more mistakes than 3.5 did.
0:16:04 And then Google came out with, like, 2.5 Pro, their Gemini model, and that’s just been amazing.
0:16:05 Like, people haven’t talked about that enough.
0:16:08 Like, Google has taken over coding with AI in the last month.
0:16:12 I think most people probably don’t realize that, but, like, that has happened in the last month.
0:16:13 Yeah, no, that’s true.
0:16:15 And, I mean, I still do quite a bit of coding.
0:16:19 I’ve kind of been using WindSurf a little more than anything, a little more than Cursor, really.
0:16:25 And, yeah, I’ve been pretty much using Gemini 2.5 the most.
0:16:29 And every once in a while I’ll get hung up and I’ll go to Claude and be like, well, can Claude solve it?
0:16:32 And sometimes Claude can solve random stuff that Gemini can’t.
0:16:32 Yeah.
0:16:42 But I would say, like, 95% of my coding now is done with Gemini 2.5, which will be interesting because now OpenAI just released 03 and 04 Mini,
0:16:53 which this new 03 model is their most powerful model, but they also have 04, which is a bigger number than 03, but not as powerful as 03.
0:16:54 Right.
0:16:57 And, you know, Sam Allman had said a few things about these models.
0:17:04 Well, more recently, he said, the benchmarks you saw for 03, we found some ways to improve upon the model more than we anticipated.
0:17:11 So it’s actually better than what we showed, but also in the past, he hinted at 03 Pro being amazing.
0:17:15 And like you said, off camera, maybe 04 is 03 Pro.
0:17:15 Maybe.
0:17:15 Yeah.
0:17:16 I’m not sure.
0:17:16 Yeah.
0:17:16 Yeah.
0:17:20 Or maybe 03 Pro might be what they launch as GPT-5.
0:17:21 Who knows?
0:17:22 Yeah.
0:17:27 Or 03 Pro could be like in a week or two and it’s only for the professional, you know, users or something like that.
0:17:28 That makes sense.
0:17:28 Yeah.
0:17:32 This 03 is their current most powerful model that we have access to.
0:17:35 It’s available to both Pro and Plus users.
0:17:41 04 Mini, I believe, is a faster, less expensive model, but still a very impressive model.
0:17:42 Yeah.
0:17:44 We can kind of look at some of the charts that they show here.
0:17:50 So they actually had another chart in their video that they’re not showing here for some reason.
0:18:03 But you can see that like they’re 04 Mini with no tools scored 92.7 on competition math and they’re 03 with no tools scored 88.9.
0:18:08 So when it comes to math, I guess 04 Mini is better than 03.
0:18:09 Yeah.
0:18:21 But then they show this other chart in their launch video, which showed 03 with Python and it got like a 99.5% or something crazy like that on the competition math.
0:18:23 Basically like acing it.
0:18:23 Yeah.
0:18:25 That’s why they had to have something like GPT-5.
0:18:31 Like they’re going to have to have like one smart router to like, because all these models are good at different things.
0:18:34 And some of it, you discover it over time.
0:18:37 Like some of it, I’m just like, you learn that like, oh, this model is just slightly better for this one thing.
0:18:39 Yeah.
0:18:39 Yeah.
0:18:47 So we can see here that like 03 and 04 Mini, like when it comes to competition code here are like pretty equal.
0:18:59 And then down here, the PhD level science questions, 03 is actually better at that than 04 Mini, you know, but 04 Mini is better at math, but not as good at science.
0:19:02 How big is the jump in general from like 01 to 03?
0:19:06 So we’ve got the estimated inference cost on this axis here.
0:19:25 And then this is the math benchmark here and we can see 03 Mini, you know, is the lowest cost, but also didn’t score as well where you’ve got 04 Mini high on math here, pretty expensive comparatively, but also, you know, really smart.
0:19:25 Right.
0:19:36 And then we can sort of see it in comparison here to 03, which is not as good, but it’s not showing like the jump from like a one pro to 04, which is kind of frustrating.
0:19:37 Yeah.
0:19:41 Well, I mean, yeah, I think, I think 03 Pro be the more comparable there for a one pro.
0:19:43 I think they’ll probably come out of that soon.
0:19:49 But then here we can see when it comes to GPQA, what does that test for?
0:19:52 Graduate level Google proof.
0:19:58 So basically graduate level questions that you can’t just Google an answer for is essentially what that means.
0:20:08 So on that benchmark, 03 Mini actually performs about the same as 04 Mini high, which is interesting.
0:20:11 They both perform the same.
0:20:13 Wait, so the GPQA is actually physics, by the way.
0:20:15 So it’s like physics questions.
0:20:17 Because I was like, yeah, hard questions.
0:20:20 So cost performance, 01 and 03.
0:20:21 Okay, so here we go.
0:20:22 We can see this is math.
0:20:25 A-I-M-E is math questions, right?
0:20:29 So we can see on math, 03 for inference costs is down here.
0:20:32 03 high is way up here.
0:20:38 And the inference cost is like way cheaper than 01 Pro, which I believe is the same as 01 high.
0:20:39 Yeah.
0:20:41 I don’t know why they’re using different terminology.
0:20:43 I think 01 high is 01 pro.
0:20:58 I mean, I think, you know, remember when we talked about when like reasoning models first came out, I think when we were out in Boston, you know, for the HubSpot inbound conference, you know, we talked about like how things were going to speed up now that there was this new paradigm where it wasn’t just throwing more data at it, but you also could throw more compute power to.
0:21:01 And so we probably would see things improve dramatically faster.
0:21:09 It feels like we’re now seeing that because instead of the model getting, you know, twice as good in like, you know, two years, it’s like now it’s like what, like in three months or something.
0:21:10 Yeah.
0:21:10 Oh yeah.
0:21:15 The, the amount of time between like each of these new releases is just like compressing.
0:21:16 This is so fast.
0:21:17 I mean, here’s one that’s really impressive.
0:21:21 That same GPQ a here, here’s the curve of Oh three.
0:21:24 Look at where Oh one pro is basically down here.
0:21:34 The cost is like the highest possible cost and it’s dumber than Oh three low model or about on par with the Oh three low model here.
0:21:42 And the Oh three high is still maybe slightly cheaper than Oh one medium, but like, look at the jump when in intelligence way up here.
0:21:43 Wow.
0:21:43 Pretty wild.
0:21:47 The speed at which these leaps are happening is pretty crazy.
0:21:48 Yeah.
0:21:50 When we first started the show, we were talking about like how big of a change.
0:21:55 It just wasn’t, I think people probably thought that was hype and it’s actually happening now.
0:21:59 And I think people still don’t understand it’s actually happening and like how big of a deal it is.
0:22:02 Like Sam Altman was interviewed at Ted the other day.
0:22:02 Oh yeah.
0:22:03 I saw that.
0:22:03 Yeah.
0:22:09 And he was talking about how, I think he said that they have 500 million weekly active users now.
0:22:17 And then the Ted guy said, backstage, you told me that your numbers doubled in the last month since the whole Ghibli thing came out.
0:22:22 He was like, that was supposed to be off record, but yeah, you know, yeah, it’s, it’s going well.
0:22:28 I mean, so possibly open the eye is what in the ballpark of maybe 700 plus million users.
0:22:31 I mean, that is just Silicon Valley has never seen this before.
0:22:38 And it’s so wild that Sam Altman used to run YC, like the big incubator for Silicon Valley, that he ends up producing the biggest startup of all time, possibly.
0:22:40 I just wonder why he would want that to be off record.
0:22:43 That’s something that would probably attract investors pretty quickly.
0:22:44 Didn’t know those numbers.
0:22:45 Yeah.
0:22:51 But I think he is in a game with Google and Grok and Anthropic and everyone of like, you want to hype things up.
0:22:54 But also you have to keep some cards secret, right?
0:22:57 Like you don’t want them to know exactly everything.
0:23:00 I mean, there’s probably some misdirection and all kinds of stuff going on.
0:23:00 Yeah.
0:23:00 Yeah.
0:23:02 It’s wild how fast it’s improving.
0:23:09 And one thing that I found interesting is in the interview, he alluded to this being possibly like the most pivotal year for AI.
0:23:13 Like basically saying, and he’s also kind of like saying the best is yet to come kind of thing.
0:23:16 And also talking about the next big thing is scientific discoveries.
0:23:16 Yeah.
0:23:21 And the biggest talk on X over the last week that I’ve seen a lot of the rumors that were going around in some of them
0:23:26 were even shared, like there was like positive things being said by scientists too, kind of alluding to this being possibly true,
0:23:34 is that they’ve had early access to the new open AI models and possibly some new discoveries have been made.
0:23:39 Or at least experiments that they were not able to come up with that are now helping them possibly discover new things in science.
0:23:40 Yeah.
0:23:45 So possibly this is like having like really, really material impact on science for the first time ever.
0:23:46 Yeah.
0:23:46 No.
0:23:48 In fact, let me pull up an article.
0:23:50 I was actually reading this earlier.
0:23:52 It’s on the information here.
0:23:56 Open AI’s latest breakthrough AI that comes up with new ideas.
0:23:56 Yeah.
0:23:59 Now AI is getting good at brainstorming.
0:24:05 This came out a few days ago before this new, you know, O3 and O4 mini came out.
0:24:09 If the upcoming models dubbed O3 and O4 mini perform the way their early testers say they do,
0:24:15 the technology might soon come up with novel ideas for AI customers on how to tackle problems such as designing
0:24:20 or discovering new types of materials or drugs that could attract fortune 500 customers,
0:24:23 such as oil and gas companies and commercial drug developers.
0:24:28 in addition to research lab scientists, the apparent improvements highlight the benefit of AI models
0:24:32 focused on reasoning, which the chat GPT maker debuted in September.
0:24:37 But yeah, basically saying that these new models, the people that got early access have been saying
0:24:43 like this is helping us design and discover new materials and drugs, which is that’s kind of the
0:24:44 holy grail of what people want AI to do.
0:24:45 Yes.
0:24:49 I mean, it looks like we’re already there and we’ve just like scratched the surface of what this new
0:24:51 paradigm with reasoning models can do.
0:24:55 But it seems like the next several years, it’s going to keep accelerating and possibly,
0:24:57 you know, revolutionizing science.
0:24:59 So it’s like just, you know, it’s exciting.
0:24:59 Yeah.
0:25:00 Yeah.
0:25:03 And one of the things about these new models that we haven’t even touched on yet is that
0:25:05 they’re all being built multimodal, right?
0:25:09 So theoretically, you can give them videos, you can give them audio, you can give them text,
0:25:13 you can give them all the modalities, images, anything.
0:25:17 And it will actually be able to understand what’s going on within those.
0:25:21 But also they all have search now and they’re all getting the ability to use tools.
0:25:24 And so this chart I found pretty interesting.
0:25:27 This is one that they showed in their live stream, but they didn’t actually put on their
0:25:27 website.
0:25:33 You can see that when they gave Oh, for many tool use and it was using Python to help with
0:25:38 the math, it scored 99.5 on the competition math.
0:25:43 So like, yeah, basically acing competition math, when you give it access to be able to
0:25:48 use tools, it’s getting to that point where like these benchmarks aren’t good enough anymore.
0:25:53 We need new benchmarks for like, can AI find new discoveries for us?
0:25:54 Like, how do we make those kinds of benchmarks?
0:25:56 Right.
0:25:58 We’re getting more and more to a situation too.
0:26:01 Like when the Oh one came out, it took people a while to realize how good it was or like what
0:26:02 it meant.
0:26:03 You know, Sam Altman said this.
0:26:08 He said, his son is never going to grow up in a time where he’s smarter than AI.
0:26:12 Like for a child who’s being born right now, who’s young, that’s just not going to be possible
0:26:13 in their lifetime.
0:26:14 So like, that’s where we’re at.
0:26:18 Like these models, not in every single way, not in an entirely general way, every single
0:26:23 possible way, but in a lot of ways, they’re already more intelligent than most people.
0:26:25 And that’s just going to continue and accelerate.
0:26:29 So I’m constantly thinking about like, what does this mean for me personally in my business?
0:26:31 And like, what am I going to, how am I going to use it?
0:26:33 I saw Greg Eisenberg tweeting about this the other day.
0:26:36 If he’s telling the real story, if he was just trying to like set up something for social
0:26:40 media, but he’s talking about how to be kind of in a rut because things are changing so fast.
0:26:42 And I feel like there’s so many different things you could do.
0:26:46 And it’s like, what do you do in an environment where things are so dynamically changing?
0:26:48 We’ve never experienced this in human history.
0:26:53 It’s not clear how anyone exactly takes advantage of this because maybe by the time you do something,
0:26:54 it’s already changed again.
0:26:54 Yeah.
0:26:55 Yeah.
0:26:58 We’re entering completely new territory that nobody’s been in yet.
0:27:04 And even all the people that are trying to predict when things are going to happen or what’s going to happen next are all just making predictions.
0:27:07 Like they don’t even necessarily know what’s coming or where this is going to lead.
0:27:08 Yeah.
0:27:12 I’ve been preaching this for a while, but like, I really feel like major corporations are
0:27:13 not taking this seriously enough.
0:27:18 Like, you know, being a corporation who has too many meetings and overly like moves very
0:27:18 slow.
0:27:20 That’s a bad spot to be in.
0:27:24 I think people don’t realize how smart Elon Musk has been in this area where he’s like
0:27:26 really like made things more lean and like move fast.
0:27:31 Every big company should be doing that because otherwise, how do you survive in an environment
0:27:33 where things are dramatically changing?
0:27:33 Yeah.
0:27:33 Yeah.
0:27:37 Did you see Toby Lucky’s statement from Shopify?
0:27:38 Yeah.
0:27:38 Yeah.
0:27:42 I mean, he’s basically saying what you said, but the big point that he made is if you guys
0:27:48 want to hire anybody else in Shopify, you must prove that what you’re hiring for can’t be
0:27:49 done with AI first.
0:27:50 Right.
0:27:54 If you can prove AI cannot do it, then okay, you can make the hire, but you must prove to
0:27:56 me that AI can’t do it first.
0:27:56 Right.
0:28:00 And it’s also super clever because he’s also kind of like baking in AI into the culture
0:28:03 of the company where it’s like, look, if you’re involved in the company, you have to be thinking
0:28:06 about how you could be using AI to make the company better.
0:28:06 Yeah.
0:28:10 Because otherwise, how could you do that test if you’re not like actually using AI and like
0:28:11 actively understanding what it can do?
0:28:13 So that’s super brilliant.
0:28:16 And I don’t think he’s coming from a place of like, AI is going to replace all of you.
0:28:20 I think he’s coming from a place of like, this company can be so much more efficient
0:28:24 and so much better and get things done so much more smoothly if we’re all leveraging
0:28:25 AI together.
0:28:28 He’s talking about like an AI first company, like even though, even though they’re a large
0:28:32 org restructuring already to be an AI first company.
0:28:34 And most companies are not realizing this.
0:28:38 I mean, I, I think almost all major companies should be reorganizing right now ahead of time,
0:28:42 but you know, like people always, you know, they’re more reactive and proactive, like especially
0:28:46 big organizations, like all the companies, they will realize this like five or 10 years from
0:28:47 now.
0:28:50 And then some of them will not survive because of that, but it’s gonna be a wild time.
0:28:51 Yeah.
0:28:53 So I want to bring this one home with this.
0:28:54 Yeah.
0:28:59 So I actually asked open AI is deep research to sort of break down all of the models,
0:29:04 like help get past the confusion, do the deep research on which models are good at what,
0:29:09 and then give me like a ranked order of basically which was the dumbest to the smartest of all of
0:29:11 the available models right now.
0:29:21 It actually took, let’s see, 16 minutes for even a chat GPT to figure out, um, its own dang models
0:29:25 with how confusing it all is, but here’s the breakdown.
0:29:32 So GPT 3.5 is great at basic conversations, but substantially lower reasoning and knowledge
0:29:32 than the others.
0:29:38 Then GPT four, which came out in March, 2023 was a big leap from 3.5.
0:29:44 And then you have GPT four turbo slash GPT four Oh, which apparently are sort of tied.
0:29:51 Then you have GPT 4.1 nano, which is one that came out this week, which is a very small, but
0:29:57 surprisingly capable model, 80% of GPT fours ability, basically.
0:30:02 Then you have a GPT 4.1 mini slash GPT four Oh.
0:30:04 So GPT four one mini was one that came out this week.
0:30:08 GPT four Oh is the one that’s sort of been the default model for a while.
0:30:15 inside of chat GPT, but GPT four one mini, it says matches or exceeds GPT four Oh’s intelligence.
0:30:17 So both are kind of tied.
0:30:24 Then you have GPT 4.5, which was the one that they codenamed Orion GPT 4.5.
0:30:26 I know it’s been available in the $200 a month plan.
0:30:30 I can’t remember if it was available in the $20 a month plan, but that one’s been one that
0:30:32 was sort of more human.
0:30:37 like, I guess it’s good at writing stories, good at when you just want to have like a
0:30:37 chat with an AI.
0:30:42 That was the one that I always felt like was the best at having like a very human like conversation.
0:30:44 Then you have open AI’s Oh one.
0:30:49 This was their first reasoning model where it actually sort of thought through after you gave it the prompt.
0:30:53 And so during inference, it sort of did that sort of chain of thought thinking.
0:30:55 This was the first model we saw that was doing that.
0:31:05 Then you have Oh four mini, which was, I believe one that also came out this week is a little bit better than just the standard Oh one.
0:31:09 So it’s a new efficient reasoning model when run in high compute mode.
0:31:15 It’s extremely capable, likely outperforming Oh one and approaching the top models in STEM benchmarks.
0:31:26 Then you have GPT 4.1 full, which is another one that came out this week, which is open AI’s latest GPT model only available in the API right now, not available in chat GPT.
0:31:37 And this one’s more tuned for coding and following instructions, but this is why it’s a little bit confusing because we can see GPT 4.1 is actually a smarter model than GPT 4.5.
0:31:42 So even though the numbering scheme went backwards, it’s actually a more powerful model.
0:31:49 And then you have open AI Oh one pro, which for the longest time was open AI’s top dog model.
0:31:54 If you got into chat GPT and you had the $200 a month plan, Oh one pro was the best of the best.
0:31:58 It just kind of crushed anything you threw at it.
0:32:02 It was really good at coding took forever to get into the API, but that was their best for a long time.
0:32:06 And then you have open AI Oh three, which was also one that was just released this week.
0:32:08 It is available in chat GPT.
0:32:13 It’s available both in pro and plus so 20 and $200 a month plans.
0:32:19 This one is their new best, most powerful model they’re making available right now.
0:32:20 That’s open AI Oh three.
0:32:28 And then this list says their most intelligent, best possible model you can use is Oh three with tool use turned on.
0:32:35 So allowing Oh three to use tools like Python and things like that to help it with its, you know, math and coding and logic and things like that.
0:32:38 When you use Oh three with tools, it’s their best model.
0:32:40 And then also GPT 4.1.
0:32:44 They’ve also sort of tied with it depending on the use case.
0:32:46 So insanely confusing.
0:32:46 Yeah.
0:32:50 But this is the order it ranked it in based on all the deep research.
0:32:50 Yes.
0:32:54 I mean, so thankfully, like people won’t have to worry about this soon.
0:32:57 So like, hopefully in like two months, you won’t have to worry about this GPT five, hopefully be out.
0:32:59 It’ll manage all that for you.
0:33:00 You just talk to one model.
0:33:05 But right now, I think their descriptions actually are pretty good in line with my experience.
0:33:11 Like, I think there’s like three models people really would the average people would use right now, like for Oh, that’s what most people would use.
0:33:13 Like I said, that model keeps getting better.
0:33:13 Yeah.
0:33:16 And if you’re just using ChatGPT free, that’s what you’re using.
0:33:17 You’re using 4.0.
0:33:17 Yeah.
0:33:18 4.0 is great.
0:33:19 It just kept getting better.
0:33:23 They haven’t really like made a big fuss about it, but it seemed to have just kept getting smarter in some way.
0:33:25 I don’t think that’s entirely reflected in the benchmarks.
0:33:28 4.5 still seems to be the best for writing.
0:33:31 So if you’re wanting to edit something or write something, 4.5 is the best.
0:33:33 That one’s also getting phased out.
0:33:34 So use it while you can.
0:33:36 Use it while you can.
0:33:38 I think it’ll still be there in five.
0:33:39 I had to guess behind the scenes.
0:33:48 And then O3, which I haven’t used yet, based on the benchmarks and my understanding how the models work, I would assume that’s going to be the best if you have any complicated questions.
0:33:52 Like if there’s something that you’re thinking like, oh, the model might give me too surface level of an answer.
0:33:54 It’s not taking the time to think it through.
0:34:00 If you’re asking something complex, you probably want to hand it to O3 and give it the time to really think through its response to you.
0:34:03 And that’s also probably the one I would imagine you’re going to go to for coding as well.
0:34:04 Yeah, probably so.
0:34:06 You know, I haven’t had a chance to test any of it with coding yet.
0:34:12 There’s one last thing that we didn’t actually bring up about the new models that they just released, and that’s the context window.
0:34:12 Yeah.
0:34:15 One of these models is actually a 1 million token context window.
0:34:17 Is it a real one?
0:34:20 Like I remember that was a part of the whole fiasco with Meta as well.
0:34:21 It was like they said like 2 million context window.
0:34:27 And then apparently at all the benchmarks, it would like start to like fall apart after like 128K context or something.
0:34:28 I don’t know.
0:34:33 I actually kind of trust OpenAI’s word on this a little bit more than I trust Meta’s word on it.
0:34:33 Yeah, me too.
0:34:34 Me too.
0:34:35 Because that was just ridiculous.
0:34:36 Like people tested it.
0:34:40 It’s like they were hyping up like 2 million contacts or 4 million or whatever it was they hyped up.
0:34:43 And then the people tested it.
0:34:47 And as soon as it went beyond like 128K contacts, like the model just like became really dumb.
0:34:49 Yeah, I heard the same thing.
0:34:55 I mean, I’m sure there’ll be some diminishing returns like O1 Pro even like, you know, that’s why I was using a repo prompt is like, yeah, it could take a lot of context.
0:35:02 But you do want to kind of be careful about what you give it because if you give it more exactly what it needs, you do tend to get better results back.
0:35:09 But yeah, these context windows are getting to a point where like context length is not going to be an issue for very much longer.
0:35:09 Yeah.
0:35:13 Like you’re going to be able to plug in entire code bases and it’s going to be able to read the whole dang thing.
0:35:20 Yeah, and it is interesting that like Sam Altman kind of admittedly, well, he’s at number one, like people were asked about like DeepSeek and he’s like, well, it hasn’t impacted our growth at all.
0:35:22 So not really too worried about it.
0:35:28 But also he was saying that like moving forward, it’s probably more about the things they build on top of the models.
0:35:31 He thinks they’re going to continue to be in the lead.
0:35:35 He said, but that lead is going to narrow and it’s going to be like a small lead in the future.
0:35:37 So it’s all about what they build on top of it.
0:35:41 You know, it did seem early on that like Open the Eye was going to be incredibly ahead of the game.
0:35:44 But now it feels like there’s going to be lots of people who are pretty close to them.
0:35:45 So here we go.
0:35:50 So as of April 2025, here’s the context window sizes for OpenAI’s latest models.
0:35:55 So it was the GPT 4.1 series, the one that’s only available in the API.
0:35:56 Yeah.
0:36:03 That one is a 1 million token context window, which includes GPT 4.1, 4.1 Mini and 4.1 Nano.
0:36:05 So those are a million tokens.
0:36:08 GPT 4.0 is still 128,000.
0:36:11 O3 Mini supports up to 200,000.
0:36:16 O3 Full, likely maintaining or exceeding 200,000.
0:36:18 But it doesn’t actually say here.
0:36:22 O4 Mini released along O3 is a reduced version of a successor.
0:36:27 Specific context window sizes for O4 Mini have not been detailed.
0:36:31 So yeah, the O3 and O4, they haven’t actually said what the context window was.
0:36:38 But the new 4.1, which is the one that’s going to replace 4.5, is the million token context window.
0:36:42 So yeah, that’s pretty much a breakdown of all the OpenAI news that came out over the last few days.
0:36:46 There was one smaller update that’s just like a quality of life thing.
0:36:53 If you are generating images inside of OpenAI, they created this library page where you can see all the images that you’ve generated in just like one spot.
0:36:56 You know, kind of cool quality of life thing that they’ve added as well.
0:36:59 Yeah, there’s another move towards killing Mid-Journey possibly.
0:37:00 Yeah, very true.
0:37:02 Yeah, you know, Mid-Journey 7 came out.
0:37:06 Like people were expecting it was going to be way better at understanding images.
0:37:08 And it’s like slightly better.
0:37:08 Yeah, yeah.
0:37:10 It is still the most beautiful model.
0:37:11 And it got slightly more beautiful.
0:37:12 Yeah.
0:37:15 It more consistently beautiful and, you know, and things like that.
0:37:20 But it seems like they did not nail the thing where, oh, the characters are going to be completely consistent.
0:37:22 Like you can like reuse the characters.
0:37:22 Yeah.
0:37:24 As far as I can tell, they did not nail that.
0:37:26 Yeah, I played around with V7 a little bit.
0:37:29 I was actually about to make a video on it and didn’t even release the video.
0:37:32 Like I only want to put out videos about stuff that I’m excited about.
0:37:36 And like I was playing with it while recording and like it wasn’t really exciting me.
0:37:38 I’m like, it doesn’t feel like that big of a leap to me.
0:37:43 Yeah, the only cool part of it was the turbo thing where you can like rapidly try ideas.
0:37:44 That’s cool.
0:37:44 Yeah.
0:37:45 And you can do voice mode.
0:37:51 And so that was actually the only cool thing about version 7 was you could just press a voice mode button and just start chatting.
0:37:54 And as you’re chatting, it starts creating the images super fast.
0:37:55 That was cool.
0:37:57 Yeah, I actually didn’t play with that mode.
0:37:58 So that’s one I’ll have to mess with a little bit.
0:37:59 That’s the actually cool part of it.
0:38:00 Like everything else is like whatever.
0:38:01 Yeah.
0:38:03 I didn’t find it that impressive.
0:38:03 Yeah.
0:38:05 Ideogram has gotten really, really good.
0:38:08 The various Leonardo models have gotten really, really good.
0:38:08 Yeah.
0:38:11 Obviously, OpenAI’s model is really, really good.
0:38:18 Like all of these models have sort of caught up to each other where like mid-journey is just sort of in the mix with them as opposed to being the leader, you know?
0:38:19 Yeah.
0:38:29 But like I said in the past, I think people don’t understand like with OpenAI’s models, you know, with the reasoning models, they’re going to be able to actually understand what’s going on in the images and the videos they create.
0:38:29 Right.
0:38:32 Like that’s going to be the unlock, just like making something beautiful.
0:38:35 I’m sure you can do that, but do you understand what the hell is in the thing you just created?
0:38:36 And can you modify that?
0:38:38 That’s the thing that’s interesting.
0:38:42 And I think, I think OpenAI is like the one who’s like by far the leader in that right now.
0:38:43 Yeah.
0:38:43 Yeah.
0:38:44 They’re on the forefront of it for sure.
0:38:45 Yeah.
0:38:52 But yeah, I mean, I think that’s a pretty good breakdown of everything that’s been happening in the world of OpenAI this week.
0:38:52 Yeah.
0:38:58 We went off on a few little tangents, but, you know, hopefully we brought it all home by sort of showing you the ranking of intelligence.
0:39:02 And I think Nathan, you did a good job of saying like, all right, these are the three models you’re probably going to use.
0:39:14 So, you know, hopefully you found this informative and helpful trying to declutter your mind on all of these different OpenAI models because it’s still confusing to us.
0:39:17 I mean, we’re paying attention to it every day and we still get confused.
0:39:19 So, you know, hopefully this helped a little bit.
0:39:23 And thank you so much for tuning in to this episode.
0:39:24 Hopefully you enjoyed it.
0:39:27 If you like it, make sure that you subscribe ideally on YouTube.
0:39:30 That’s where we’re really trying to primarily grow the show.
0:39:35 But if you prefer audio, we’re available on Spotify and iTunes and wherever you listen to podcasts.
0:39:38 Thank you once again for tuning in and hopefully we’ll see you in the next one.
0:39:39 Goodbye.
0:40:01 Bye.
Episode 55: Confused about all the new OpenAI model names like 4.5, 4.1, o3, 04-mini, and the new “memory” feature? Matt Wolfe (https://x.com/mreflow) and Nathan Lands (https://x.com/NathanLands) are here to demystify the whirlwind of recent ChatGPT updates so you know exactly what matters and how to use the smartest AI for your needs.
In this episode, Matt and Nathan break down the latest OpenAI announcements—what the new “memory” feature actually does, how it could make ChatGPT your most personal assistant, and how each of the new models stacks up. They dissect the confusing model lineup, explain what’s getting sunset, reveal how to match the right model to your workflow (from writing to coding), and discuss rumors of OpenAI’s potential entry into social media. By the end, you’ll know which model is “dumbest to smartest,” what’s coming next, and how these rapid-fire advances might reshape tech and business faster than anyone expected.
Check out The Next Wave YouTube Channel if you want to see Matt and Nathan on screen: https://lnk.to/thenextwavepd
—
Show Notes:
- (00:00) Demystifying OpenAI’s Recent Launches
- (08:51) Personalized Social Network Plans
- (11:02) GPT5 Delay and Concerns
- (15:00) SWE Bench and External Comparisons
- (16:12) Gemini 2.5 Preferred for Coding
- (22:03) Startup’s User Surge Stuns Silicon Valley
- (23:57) AI Advances in Brainstorming
- (28:54) Ranking AI Models by Capability
- (31:05) Advanced AI Models Overview
- (35:13) Sam Altman on OpenAI’s Future Lead
- (36:23) OpenAI’s Context Window Updates
—
Mentions:
- OpenAI Reasoning Models: https://platform.openai.com/docs/guides/reasoning?api-mode=responses
- Gemini 2.5: https://deepmind.google/technologies/gemini/pro/
- Claude: https://claude.ai/
- Midjourney v7: https://www.midjourney.com/updates/v7-alpha
- Sam Altman at TED 2025: https://www.ted.com/talks/sam_altman_openai_s_sam_altman_talks_chatgpt_ai_agents_and_superintelligence_live_at_ted2025?language=en/
Get the guide to build your own Custom GPT: https://clickhubspot.com/tnw
—
Check Out Matt’s Stuff:
• Future Tools – https://futuretools.beehiiv.com/
• Blog – https://www.mattwolfe.com/
• YouTube- https://www.youtube.com/@mreflow
—
Check Out Nathan’s Stuff:
- Newsletter: https://news.lore.com/
- Blog – https://lore.com/
The Next Wave is a HubSpot Original Podcast // Brought to you by Hubspot Media // Production by Darren Clarke // Editing by Ezra Bakker Trupiano