Claude Opus 5.5 is SCARY GOOD!

Julian Goldie (0:00)

Opus 5.5 just dropped from Claude, and it is very impressive so far. I've been testing all day, running loads of different tests, and from everything that I've seen so far, it is very impressive. Now, one of the things that you need to know is that Claude Fable 5.1 is pretty much on par with Opus 5 So at this point, you might as well switch over to Opus 5.5. And bear in mind, like, Opus 5 itself from Claude was pretty disappointed. So I didn't have high hopes for 5.5, but when it dropped today, actually is very, very good. Now, if you're wondering how to use it, you can get it inside Claude. You can just switch the model over to Opus 5.5, as you can see right here. Just make sure you've gone to Claude and then check for updates to switch over. And then once you've done that, you can start using it straight away pretty much. Now, if we look at the benchmarks here, it's absolutely crushing 5.1 from Fable, and it's also beating Opus 5, and it's actually beating GPT-6 Astra on some tests, although test out yourself. I wouldn't take that at, you know, face value. And then also you can see here it's beating GPT-56 Sol, which I would expect. But yeah, pretty much a frontier model. The difference here, of course, the is a lot, lot cheaper and you get way more tokens. Like normally I'm on the usage credits for Fable 5.1, and we get charged for it. With Opus 5.5, we're not getting charged for it. You know, it's well within the usage limits. If I open this up, for example, you can see no one near the five hour limit for weekly all models and weekly Fable. This is the problem. You see that there's a limit for Fable 5, which is super annoying and that goes up very, very quickly. Whereas, for example, if you're using Opus 5, well, then you get way more tokens on that plan, which is absolutely awesome. Now, it's also designed to run on longer tasks as well, which is good. You see how well it performs over here. For me personally, I've been using it on medium. So you can switch the effort levels here. But for me personally, I think like just stick on medium is probably enough to get good outputs. That's actually how we build the tests on Goldiebench that I'm going to show you in a second ago. So we've been testing out all days, you can see right here, and it builds some pretty impressive stuff, as you can see, right? Pretty fun. All the games are smooth and also they all works like first time round, which is pretty amazing. And if we open up the stuff, I mean, some people say, why are you testing out games, Julian? Why not test it for AI automations? And I'll show you an example of like AI automations in a second. But from what I'm actually seeing, you know, it's really good for testing with games simply because it's visual and also it gives you an idea of the logic of Opus 5.5 as well, right? So how good is it at planning? How good is it at reasoning? You can see this kind of like GTA style game that we're playing right here is kind of like a racing game. But it's just so much fun and there's so much crazy stuff and like the graphics and everything else is super nice too. This is like a open world game. And like if you just look at the quality of it, it's super nice, it's easy to use, it's smooth. Again, this was built first time around. We've got like this little giant enemy over here. And you can see like a whole open world is absolutely beautiful to use and to play with and to test out. So if you haven't already, definitely check out Opus 5.5. And I think you'll be blown away by how good it is. I didn't have high hopes for it at all. Some things look a little bit basic when it comes to details. You can see right here. But like the vibe and the style of it that creates this is pretty good. It's pretty impressive. So let's have a look at some more stuff here.

We'll just keep going through. Actually, the parachute drop game was pretty fun here, as you can see, so we can just parachute away. And also, if you didn't realize already, there are actually a few other things going on right now. So, for example, we had GPT-6, Luna, and Sol just drop today.

But the biggest differences with those is like, they're not quite as exciting. I mean, you will get a lot more usage out of something like, for example, Sol, which is still Frontier, but it's just not as good as Astra. By the way, this is an example of Opus 5.5 using video animation. So it's got the re-motion skill here. It's got the progress bar at the bottom. And you can see, let's create this full animated beautiful video, as you can see right here. By the way, if you want to learn how to use this stuff, link in the comment subscription or go to the aiprofitboram.com. We've got loads of training, systems, everything you need to win with AI automation in Opus 5.5. So overall, very impressive model.