Qwen 3.8 VS VS Fable 5: Who Wins?

**Julian Goldie** (0:00)
Qwen 3.8 versus Fable 5, who wins, actually built over 50 different tasks with Qwen 3.8. And I'll show you what we discovered and how they compare in a second ago. And also, if we compare them on the benchmarks here, you can see the Fable 5 is on a similar level on some of the benchmarks versus Qwen 3.8. But the difference is the Qwen 3.8 Max is actually free, number one, and number two, it's actually going open source as well. So this is number one, like a huge threat directly to Fable 5 And I think they probably got to release something soon, especially with Opus 5 catching up. But at the same time as well, I want to show you how it actually performs in reality, because benchmarks and actual tests are very, very different. So if we have a look, for example, on Terminal Bench 2.1 for Terminal Agent, you can see here that Qwen 3.8 Max is scoring 86.6 versus Fables 8.6, right? So there's actually quite a big difference. And you can see here that in some ways, Qwen 3.8 Max is actually outperforming all other frontier models on the market. How does it perform in reality? Well, that's the most important thing, right? So if we have a look, for example, all right, so if you compare these side by side, you can see our live builds. So these are Qwen 3.8 versus GPC 5.6 Sol versus Fable 5 versus Kimi K3. Now, first thing you're going to see is this generation from Qwen 3.8, super nice, like SIM 3D, it feels nice to use, easy to use, and it's actually a step above all the other builds. Now, when we've tested this, this doesn't mean that Fable 5 straight away is not as good as Qwen 3.8, but you can see the full power of Qwen 3.8 and how good it actually is when it comes to building stuff. Here's another example. So this is a kind of open world RPG game, as you can see, and the gameplay is smooth, it's pretty easy to use, pretty fun to play, pretty difficult and challenging as well to play as well.
And if we compare that versus, for example, Fable 5, Fable 5 is like good, but it's a little bit slower, not quite as interesting straight off the bat, and also bugs out a little bit when you're moving forward here. Now, Kimi K3 can also build super nice stuff. So if you actually look on front end design, I would say that the front end of Kimi K3 in this example looks way nicer than the front end of Qwen 3.8 Max. Some other stuff like this Neon Racer game, again, like huge step up versus Fable 5 on this particular design, which is pretty nice. And I just noticed that everything feels super fluid and 3D when you're building out with Qwen 3.8 as well.
Here's another example of like a 3D racer game.
This actually, I wouldn't say is that good just because it's a bit difficult to play once you get past that bit. Pretty fun. By the way, if you want to check out these builds, you can check them out at goldiebench.com. We have all of our builds and all of our tests side by side with the leaderboard so you can see which model you prefer and also play the tests and see what you think yourself, or you can test them out yourself with similar prompts. The thing that I would say here as well is like it's really good to test out simply because it gives you an idea of coding, logic, how it plans, front end. It gives you basically everything in one single test, which is super useful. Now, if we have a look at this RPG game as well, pretty good. I mean, the camera angle is not that great, but again, really, really nicer. Let's check out the flight simulator. Now, usually I've noticed for some reason, like Kimi K3 and open source models in general struggle with this. Now, if we have a look at Fable 5, Fable 5's version is probably the nicest, which is this one down here.
You can see with Qwen 3.8 that it's really struggled on the camera angle. It doesn't really make sense. The controls are weird, and you can't really use it properly. You can't play it really well. So that's the first time I've seen Fable 5 actually outperform Qwen 3.8 on all our tests. Let's have a look at this one. Now, this is really cool. Like the intro, as you can see, the pop-ups, it's a super fun game to play. Pretty crazy, actually, once you start playing it. And the detail as well on it is really, really nice. If we have a look at Fable 5, what you can see over here is it's not so good. The other thing that I would say that's a big disadvantage with Fable 5 is sometimes it will just automatically switch over to Opus, which means you can be halfway through a build and then you're like, well, why am I getting bad results? And then you look and you're like, oh, it's because it automatically switched over to Opus instead of continuing to use Fable 5, which blows my mind, really, because you could be coding something out for hours, it could be running on goal mode, and then it just automatically switches over to Fable 5 And this is also a lot slower with less detail to play, whereas the version with Qwen 3.8, much more interesting. So I'm going to look at this Dragon Realm game here.