NEW Claude Opus 4.6 DESTROYS Codex 5.3?

Claude Opus 4. 6 versus GPT codeex 5. 3 who wins. Today we're going to be testing them out. So both of these new updates, one from Anthropic and one from OpenAI were both released at the same time, right? And you can see the benchmarks here. GPT3 5. 3 is nowhere near on the benchmarks because it was just released literally at the same time OPUS 6 was released. So you can see here this is GPT 5. 3 C codeex and on the benchmarks codeex actually wins. But what we're going to do today is we're going to test them side by side and see who performs the best, which one creates the best content. We got B here who's first on the comments. Shout out to B. And we're going to get straight into this. All right, so I'm going to use

some prompts from the AR success lab which is a free community link in the comments description or just ask our GPT for Julian Goldie's AR success lab. connects you 52,000 members. We'll be testing out directly some of these prompts to see which one performs the best. All right, so we're going to grab this one like so. We are going to go over to the chat inside open router and then we can compare them side by side. All right, so we're going to use 4. 6 over here and then we'll use codeex 5. 3 if it's inside there, which it doesn't seem to be. So what we'll do instead is we will run them side by side inside codeex and opus. So let's do that instead. So we got codeex on the right hand side and then we have opus on the left hand side.

So let's plug in this prompt. We're just going to switch over here to 4. 6. There we go. And we'll run the same prompt inside both and see which one performs the best. So you can see they're beginning to uh to code out. This is for the first test which is build a hyperdopamine pong game. So we'll see how they perform. What we got on the questions here? So we got G who says but about Gemini 3. 0 Pro. Honestly, I don't think it's in the race at this point. Victor says, "Hey, good to see you here. " Whilst we're waiting for those in the meantime, let's see how they compare on the actual stats. So if we go to models over here and we'll have a look at Opus 4. 6. six and you can see it does have a million context window. These are some of the things that have been developed by uh codeex 5. 3. Actually compare, we can see GPT 5. 2 versus 5. 3 looks very similar to be honest in the way that it's coded out. She says, "Oh, okay. Glad we agree. " And it looks like 5. 3 has finished first just by a marginal fraction. But we have the preview of Claude Opus 4. 6 here. So, let's try this out. Just make sure this is muted. Hyperpunk synth wave version. There we go. So, you have to control both sides, which doesn't make any sense. All right. It is a bit mad. I don't know why it's made it so that I have to play both sides. So, it doesn't make sense. Should be me against the computer, right? All right. So, that is Opus. Looks pretty cool. I'm just going to switch over to the code cuz it's going crazy. Let's have a look at the GPT codeex one. We'll say open this up. Should be able to use this inside a browser. Here we go. So, this is the one from GPT 5. 3. See, like that's actually got us playing against each other. It's not very smooth, though. If you look at this one over here, look how nons smooth this is. I would say the Claude Opus one was a lot more exciting, but at the same time, this is a smarter move because there's no way on a laptop you're going to play like uh two different people playing each other, right? So, that didn't really make sense for Opus. So, I'm going to say that Codeex one just about purely because number one, it's a nice game, but number two, you've actually got a playable character on each side, which is better, but both of them are pretty nice. All right. So, chat GPT codeex 5. 31 opus 4. 6 zero. Let's try something else now. So

we're going to try creating a space invaders game. So, we're going to take that. We'll go back to both of these. When are we going to see something useful? Rap says where they can consume blueprints and build out the back end in GitHub. Actually, I mean like they can open GitHubs pretty easily right now. So for example like earlier today I got claw code to install open claw and the same with chatch 5. 3 actually for that test uh claw code one with opus 4. 6 whereas with 5. 3 it couldn't figure out how to use 5. 3 inside openclaw and the whole thing broke. So um for that individual test for loading a GitHub, we actually saw that Claude beat Chachi PT5. 3 took them both ages to set up by the way. Saw some pretty crazy stuff on X as well. So for example, like people one-shot in multiplayer games. Look at that. How crazy is that, right? But some of them take like hours to create. So for example, I think someone actually created Pokémon with Opus 4. 6. It looks super impressive, but it took like 8 hours. You see another one here side by side. So this is a tweet by Venet and you can see that you've got Opus 4. 6 versus 5. 3. Both of them one shoted it which is pretty crazy. So let's go back here. So we got the game ready. 5. 3 is still thinking. Opus has already finished. Let's try this out. Pretty nice. Nice game. Let's try out this. So I'm going to open this up. And here we've got 4. 5. It's got sound effects as well. Honestly, I would still say that Opus 4. 6 six didn't perform as well. The game was less playable versus the game from GPT Codeex 5. 3. I'll be interested to know, let me know in the comments which one you think is better. So, this is Opus 4. 5, uh, sorry, 4. 6, this one over here. And this is Chat GPT 5. 3, right? Or GPT codeex 5. 3. But you can see it's less playable this one. like it's a lot harder whereas this one you can play for a lot longer and it just feels nicer to use. The other thing that I was going to

say here that's quite interesting is if you are using Claude code this is something that I've only seen Claude be able to do so far. You can do something called agent teams, right? And what that means I've got a full guide inside the air profit board here. What that means essentially is you can conduct like one task and give it to multiple agents to work on in parallel. Right? So you see a bunch of use cases for this. So, for example, you can say like, I've got 20 video scripts. Here's what I want you to do. I've got a piece of content. Right? Let me show you an example. So, if we take this one and we'll say, create an agent to analyze my top 10 performing videos, spawn three teammates, blah blah, have them debate, and create a playbook document. And then we'll say, actually, just do one video script. see below. And I'll just do this as a demo. So if we take that copy like so, plug it into cloud code and we just have to make sure we're using claude and then you can create a team of agents to work together which is pretty cool. The only problem with that is that if you've got multiple agents working on the same task, the biggest issue with that is that it will use up a lot of tokens, right? And it's not a cheap API as most people know watching this. So that's something to watch out for. But it's a unique feature that Only Claude has. Now overall, I would say that I am going with

I would still say I'd prefer Claude overall. Right. The reasons for this, number one, Claude code is just so much nicer to use than Codeexos. Number two, you can use agent teams, which I think could be really powerful. So, for example, I could have multiple agents creating multiple thumbnails at the same time. And number three, I just think that Claude is still like a lot more sentient, particularly if you're using something like open claw or claw bar. So, overall, I'm going to go with Claude personally, but I'd be interested to know inside the comments like do you prefer chat GBT codeex? Are you switching to Claude Opus 4. 6? Are you using either of them or just going to stick with some of the older models that are cheaper, etc. Be very interested to know. So, thanks very much for watching. If you want to get access to everything that I've covered today, you can get the video notes for the comparison tests inside the AI success lab link in the comments description. This connects you with 52,100 members. And if you just type in the top like this, you can get access to all the prompts and the templates that I tested today for free. All right, along with all my training as you can see on the left hand side right here. Now, if you

want to get my AI automation community that shows you how to save time, scale with AI and grow, then you can check out the AI profit boardroom. This is an amazing community where you can ask questions, get help, get support whenever you need it. You can share your wins, etc. And also inside here, you can jump on four weekly coaching calls a week. So, you can jump on these coaching calls, get live help on a Zoom call. And also, if you miss them, you can watch them back inside the classroom right here. Additionally, you can learn how to go from complete beginner to expert with AI in just 5 weeks. Plus, you'll learn how to build your first AI agent in under 5 minutes. On top of that, you'll get my best playbooks for AI avatar videos, how I automate shorts, how I automate my emails, how I automate Instagram, etc. And that's all inside the AI profitable boardroom along with my agency course on how to get more clients and how to rank number one with AI SEO and how to grow a YouTube channel from scratch. So, feel free to get that link in the comments description. Appreciate you watching and I'll see you on the next one. Cheers.