I Tested Qwen 3.8 So You Don't Have To…

So there was a bold statement from Gwen 3.8 from Alibaba yesterday that it's launching and going open way soon, which is pretty exciting stuff. Massive 2.4 trillion parameters, and I've tested it across loads of different builds. I'm going to show you how it compares to Fable 5 side-by-side and also GPT 5.6 side-by-side. Now you can see here that they said, we believe it's one of those powerful models available today, compatible to leading frontier AI models, second only to Fable 5.

And I've already tested it by Cuoda or Coda, however you want to pronounce it. We've built a ton of stuff with it, as you can see. So let's just get straight into the tests over here. So what we have is a game for racing.

This is the Fable 5 output on the right-hand side. And this is the output from Gwen 3.8. And you can see here that this version is clearly nicer. It's more fun to play.

I really like realistic 3D style game, looks super nice. The graphics are awesome. If we have a look at Fable 5, it's just not even in the same category. Now, does this mean that Fable 5 is worse than Gwen 3.8?

Absolutely not. We're going to be doing loads more tests side-by-side and you can see what you think. If you're wondering, okay, how does it compare versus GPT 5.6? So we actually did this little fireworks test and we've tested loads of different stuff, which I'll walk you through today.

But you can see here side-by-side that this is Gwen 3.8, GPT 5.6. So I do think that the output from GPT 5.6 sole here is much nicer in terms of the UI and the front ends, and that's actually a trend I've seen over all of the tests that I've done. By the way, quick one. If you want to learn more about AI automation, how to scale your business with AI automation, save time and get access to all of my best trainings in the Mason community, you can get that inside the AI Profit Boardroom, link in the comments description or go to the AIprofitboardroom.com.

Let's get back into it. So if we have a look, this is an Aurora test. And if you look at them side-by-side, I would still say that there's something more classy and elegant about the designs from Fable 5. They look very similar.

I don't think that's a great way to test them side-by-side. Here's a black hole simulation. You see Fable 5 is kind of weird and off. Like it just went into the bottom left.

Whereas this one looks cleaner, nicer. Nice animations, neatly organized, et cetera. So look at some of this stuff. I'm not too fussed about, I mean, these are pretty similar side-by-side.

You can zoom in and zoom out. But again, I would just say there's something about Fable 5's design that looks nicer. Looks cleaner, feels more elegant. Having said that, they've both done a good job on that.

Now this is a neon blaster test that we did. And the one from Quent 3.8 is looking way nicer. It's more fun to use, better gameplay. It's in 3D.

Fable 5 doesn't really compare, but I think if we gave Fable 5 another shot and gave it an example, it would do better. When it comes to the cloth simulation, you can see that actually the style on Fable 5 is much nicer. Like it's much cleaner to move around. This one doesn't seem to be so easy to navigate and you can't move the cloth around, whereas with this one, you actually can, and the graphics and the physics are much nicer from Fable 5, but it is, you can see it's kind of in the same category.

It's just not as good as Fable 5, I would say. This one looks super nice from Quent 3.8, and this is a neon racer game. Again, like, I mean, this one is pretty cool. Like 3D games, it seems to do really well with Quent 3.8.

The gameplay is fun. The graphics are nice. The design is super nice. Fable 5 is a little bit more retro in the style.

Also, you might be wondering, okay, how does this compare to something like GPT 5.6? So I'd still say Quent 3.8 held its own in comparison. It feels really nice to use. It's a lot of fun to play.

Yeah, I like the colors and everything like that. Look really, really good. Now we've got a flight simulator. So this is GPT 5.6 versus Quent 3.8, and it seems to be going backwards, which is weird.

Let's just reset. We just can't change the camera angle. I think that's what the problem is here. Whereas if we have a look at GPT 5.6, it doesn't look amazing.

Like the detail is not absolutely amazing, but it actually works and you can take off and everything else. So I would say GPT 5.6 won on this one. This is an interesting one. So this is a video side-by-side using Reemotion.

And look at like the font and the style of design. This one feels very much like AI. It's got those sort of typical blue neon types. It's got the icon, the logo from GPT 5.6, which just feels very AI.

Whereas if you look at this side-by-side, it feels a bit different. It's not using the generic sort of AI slop style. Let's have a look at the promo versus Fable 5. But I do, yeah, for sure.

You can see that Fable 5 has absolutely crushed that test. Like it just looks nicer. It feels smoother. The timing is better.

You can flick between the different timings here and it looks a lot more elegant in the style. But this one did okay. So I can see what they're saying in terms of like Fable 5 is still better than Quen 3.8. Now we've got a Dragon Realm game.

So this is like an open world RPG. And if you look at this, like the graphics are really simple. You know, it's got that generic look, sort of circular look on all the graphics from Quen 3.8. If we have a look at the version from GPT 5.6, it's more fun.

I like the style, you know, where it's like first person, a bit more detail to it. This one, I just, I don't see myself playing as much. Whereas this one we could actually develop into a game and I think it would look really good from GPT 5.6. Let's check out Fable 5s.

So Fable 5s as well. It's like, it just feels smoother. Whereas this one's a bit more robotic from Quen 3.8. Then we have a kind of GTA style game and this one actually impressed me a lot.

So you can see here, for example, we can move around, we can change the camera angle. Um, it's not far off. I mean, like if you gave it a few more hours to create something amazing here and added better graphics, Quen 3.8 did a really good job. Like pretty cool.

Whereas the one from Fable 5, like the colors are way off and it just feels very robotic, almost like the Dragon Realm game from Quen 3.8. Then we have Doom and this is a weird one when I test it out. So you can see Doom here and we can scroll around the map. We can have a look and see what's going on.

Um, but the colors are really weird. Like it hasn't really thought out the colors very well. It's not glitchy at all though. And I like the fact that it changes colors once you go inside.

Um, but it's very hard to play, very hard to understand what's going on here. However, it's an interesting take. It's just like, look at that, like it seems super buggy with the lighting and everything else. The lighting is, is not great on the Quen 3.8 version.

Now let's have a look at the version from Fable 5. This is more playable. It might be more basic in terms of the graphics, but it's much more playable, much more fun to navigate. Colors are much better.

Let's compare GPT 5.6. So again, I would say this is more playable from GPT 5.6. So it's almost like the gameplay is more interesting from Quen 3.8, but the graphics and the UI just seem to let it down quite often. That's what it struggles with.

It's not, I don't think it's going to get anywhere near the top of the front encoding tests compared to something like, for example, KimiK3, which creates much nicer stuff overall. This was a cool little dragon flight game. Um, actually looks way, way nicer with Quen 3.8. It looks so much nicer, so much more fun to play.

This kind of looks like an old school, uh, PlayStation game almost. Like the colors are nice. The vibe is good. The gameplay is good.

It is smooth. All the graphics and the physics work on it as well. It's actually quite fun to play that. As you have a look at the version from GPT 5.6, I like the fire, uh, action, but it's, it's a little bit more basic.

Like there's less details inside the game. Let's compare it versus Fable 5. Again, sort of super basic, not that interesting. They could do better on this.

And then we have this Nordic Crypt style game, which I thought was really good from, uh, from Quen 3.8, like it looks super nice. Fun. As soon as you start playing that you can collect items, the graphics that appear on the page are nicer. That sort of first person style works really well with this.

It does get a little bit boring after you've completed the enemies at the start. But apart from that, it's pretty good. Let's see if we can go through there. I mean, even like the lighting here is super nice.

And then if we walk through what happens, it just resets, but yeah, pretty good. Pretty good. Pretty good. Let's have a look at the Fable 5 version.

So you see how the, the controls here are not quite right on the Fable 5 version. You couldn't really play this. Like you wouldn't enjoy playing. Let's have a look at the version from GPT 5.6.

See, this is super nice. Nice graphics, weird enemies. Like if you look at that, the details are pretty bad there and we can't control it or fight with this guy. Yeah.

I would say Quen 3.8 did a better job overall, but yeah, so far impressive model. Is it Frontier Fable 5 level? I, from the tests, it can create really good stuff. I mean, like this is, you've probably seen from the demo.

So it's created like some of the best side-by-side comparisons versus all. It's definitely held its own as a Frontier model. The other thing that I would say here is it's a huge, like ridiculous step up from Quen 3.7, which was okay, but it was never really a model that you would take seriously. I know a lot of people are saying, and just want to be a hundred percent clear with you, a lot of people are saying that it's not as good in any way.

I'm just going to show you my tests and then you can judge for yourself and see what you think. So you might try it yourself and be like, ah, it's pretty average or it's not particularly good. Um, some people saying it absolutely cooked. So Omer, the vibe coder, he put them side-by-side and had a similar experience, which was, uh, Quen actually, um, didn't perform anywhere near the same level.

And actually he said it doesn't come close to GPT 5.6 or Opus 4.8. So it didn't cook. It got absolutely cooked. Sorry, just to be clear, I misread that at first.

So yeah, it's interesting. I think a lot of people will be disappointed simply because it's such a bold statement to say, isn't it? To say like, ah, this is Fable 5 level. You expect something that's Kimi K3 level at that point.

He might also say, okay, should you choose Kimi K3 or Quen 3.8, which is the best open source Chinese model? For me personally, I would still stick with something like Kimi K3 if I had the choice between them. It's actually quite difficult to access Quen 3.8, um, so far. And I mean, like, look at the outputs and the difference between them.

So if we have a look, for example, at Dragon Realm from K3, like it feels very open world, the colors are beautiful. The graphics of the snow falling down is nice. It's got lots of nice detail and the vibe is super nice on the, feeling the colors. Whereas you look at this one, it just feels a little bit more basic, a little bit more, um, less fun, more like, ah, this is just a basic game created by an AI, right?

Like certainly the graphics and the detail. So for me personally, I would go with Kimi K3 if I was building something big out. Um, I think that a good combination would actually be using Claude Code Claude Code with Kimi K3 inside it. Here's another example where Kimi K3 really outperformed Quen 3.7.

So this is like a Skyrim style game as well. And this one is cool. Like it is good. I mean, it feels smooth.

It's nice to use, um, resets very quickly. So that was from Quen 3.8, this one. And then if we have a look at the, even like the introduction video from Kimi K3 feels nicer. You're like, oh, this is going to be cool to play.

And then you start playing it and you're like, yeah, it's pretty good. Like look at the mountains, the sun, the colors, the UI. I mean, I can see why Kimi K3 tops the front end coding benchmarks. So I think a lot of people are going to put out there that they're disappointed with Quen 3.8, particularly after that massive tweet that they posted.

For me personally, I think it's a good model. I don't think it's up there with Kimi K3. I do think it holds its own with other frontier models, but Kimi K3 was far more impressive for me when it first came out. Uh, also I would say that it's quite difficult to access Quen 3.8.

So the only way that really worked for me was using Coda. The other thing worked for me. So Alibaba's token plant, you have to fiddle in a lot of details there and it's kind of messy. So if you want to get access to this early, I'll go with Coda.

Bear in mind, like if you're, for example, in the UK, which I am right now, it actually says that Quen is not available inside our region, so we can't actually go to quen.com. Uh, also here's another visual example. So this is Kimi K3. Quite nice colors, good vibe, shooting stars, et cetera.

You can easily control it. Whereas you look at the output from Quen 3.8 and you're like, this is so basic. So worth checking out, probably not going to blow you away, especially after the marketing that came out for it, but it is worth trying. Kimi K3 probably crashes, Fable 5 crashes.

Those two models are the ones that I'm paying attention to. GPT 5.6 Sol is very interesting, but I'd still prefer to use Fable 5 or Kimi K3. So thanks for watching. If you want to get training on all of this, we have loads of trainings on Quen, Kimi K3 and Fable 5 and GPT 5.6 inside the iProfit boardroom.

This is my own AI community that helps you save time and grow with AI automation. We actually have over 200 pages of testimonials and wins from people learning and growing. So it's an awesome community, lots of positivity, lots of awesome people inside there as well. And you can post inside the community.

I answer these questions every single day. Personally, inside the classroom, you get access to all of our best trainings. You can get new daily updates over here as well. And then also on the calendar, you can jump on four-weekly coaching calls, get help and support on your time.

Inside the map, you can meet people in your local area who are building with AI agents like you. And that's all available link in the comments description or go to the AIprofit1.com. Thanks for watching.