**Julian Goldie** (0:00)
Qwen 3.8 vs GPT-56 Soul vs Fable 5 vs Kimi K3, who wins? We've got 23 different builds to test out side by side today. Let me be walking through exactly how this performs. This is based on Qwen 3.8's post yesterday, so they were talking about launching and going open-weight soon. Massive 2.4 trillion parameter model. Also, this is, according to them, one of the most powerful AI models available today, compatible to leading frontier AI models, second only to Fable 5 So far from what I've seen on the test, you'll see some examples right here. It is looking really good. For example, if we have a look at this, look at this flight simulator and the quality and the details, and also the gameplay and how smooth it is, like on this flight simulator. We've got, for example, GPT-56 Soul over here, which is not too bad. This is a pretty cool game to play. I think Kimi K3 actually totally failed on this task. And then Fable 5 really got stuck on the details here. Like it just didn't create anything as nice. However, would I say that it's as good as Fable 5? Would I prefer to use it? I'm gonna come on to that later, but let's go through the builds first. So first of all, we've got this game called Outrun, as you can see right here, it's really fun on the top left from Qwen 3.8. If we have a look, for example, at Fable 5, GPT-56, even Kimi K3, they're all pretty decent. I would say GPT-56 Soul probably came out worst right there. Fable 5 is literally my favorite model as of today. And we'll come on to the final verdict in a minute. And it can create some pretty nice stuff. However, what I will say is that Qwen 3.8 by far created the best output here. It looks the nicest, most fun, most interesting. Let's move on to the Skyrim example here. So we've got the version from GPT-56. I would probably say that is the weakest out of all the generations here. You can see, for example, if we scroll through this world, even the controls are weird. So when I try to go forwards, it actually takes me backwards. Doesn't make sense. If we have a look at Fable 5, Fable 5 really nice. Like the quality, how smooth it is, how interesting it is to play, even the little details like the compass at the top. Super nice. If we have a look at Kimi K3, I mean, for me personally, I would say Kimi K3 is probably second only to Fable 5 When you look at this example, and then we can have a look at the example from the Skyrim game for Qwen 3.8. I mean, the gameplay is crazy. Like the gameplay is crazy. The graphics not quite as nice. The details not quite as nice. But is it a fun game to play? Is it pretty cool? I mean, do the graphics look great? Absolutely, absolutely. We can go full screen on any of these. By the way, if you want to check out yourself, it's all available at Goldy. And also if you want training on Fable 5, GPT-56 or Qwen or even Kimi K3, we have master classes on all of them inside the AI Profile. This is my best community for AI automation. And also inside the community, you can ask questions. I personally answer everything one, every single day. Inside the classroom, you get access to all my best trainings. You can drop all weekly coaching calls and you get all my best trainings on the models we're covering today. So let's get straight back into it.
We have a look here. Yeah, I mean, I'm not gonna lie here. A lot of people say, oh, you know, Qwen 3.8 is not quite as good or it got cooked by Fable 5 From my tests, I will say that it takes a bit of back and forth to get outputs like this. But from my tests, it holds its own. That's what I would say. Would I say it's as good as Fable 5? Do I prefer it? No. But does it hold its own? Is it still really good quality stuff? Can it build really amazing stuff? Absolutely. And I was actually testing it through QODA, so Coda, which is pretty cool. I couldn't get the international API to work, so I had to go through Coda. And also the other thing that I found was that when you are trying to access quen.com, depending on where you are, it's regional. So I couldn't access it where I was based in the UK. If we have a look here at the quality of the output, so GPT-56 Soul comes in second, I would say, on all these outputs. Fable 5 is coming in last. And then we've got, for example, the output from Kimi K3, which is just as strong as GPT-56. But I mean, Qwen 3.8's creativity here, the 3D models, everything about this game is super fun. So a lot of these games, like it is winning, it is leading the way. I mean, these are really, really cool examples. It creates something that's smooth. I always notice with Qwen 3.8, it just feels smooth when you're using it. So there's no problems with the controls. There's no problems with the navigation. The UI is never quite as good in terms of like, feeling the vibe or the ambience. I don't know how to explain it, but like, I'm not a UI expert, but I will say that I just generally prefer Fable 5 when it comes to UI. Here's another example. This was a racer game here.
11 more minutes of transcript below
Try it now — copy, paste, done:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090
Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000777645997