I Tested Qwen 3.8 Max So You Don't Have to...

When 3.8 Max just dropped, it's free to try and it's one of the best AI models I've ever tested. Now for me, I don't trust benchmarks. I actually built 45 real demos with it, games you can actually play, a website that looks genuinely beautiful, a full working operating system, and even a complete promo video all built in hours. Some builds are incredible, some were terrible, and a few beat the number one AI model on the planet, which I did not see coming.

So stick with me because later I'll show you the one move that turns all of this into your own 24-7 AI employee working whilst you sleep. I tested out Quen 3.8 Max so you don't have to. So I actually built out 45 live demos I'm going to show you today, just walk you through, you can see for yourself whether it's good or not. Now if you've seen the benchmarks already from Quen, you know that it's right up there with Fable 5 on the benchmarks, but this is why I actually created GoldieBench because I think it's important to test this stuff out yourself and quite often reality is very different.

So let's have a look through and see what we've built and see how it performs. So this was one of the first ones, like a 3D game as you can see. It actually looks really nice, it's smooth, it's easy to use, it's got good gameplay, like even the shadow effects and everything else on the screen looks super nice. Again some people will be like well why are you testing this out with games?

Games give you a good idea of the UI, the gameplay, the style, how it looks, how interesting it can be and also how well it reasons right. And you know if you test out for example building out a game with a local model often they look terrible, which gives you a good feel for like how the actual model performs in reality and also when you build this sort of stuff out it gives you a sort of idea of speed as well. Now this is another one, this is a sort of RPG game. This one I think is super basic when I look at it, like it just doesn't feel that nice and you can see here like the gameplay is not so good, the graphics and everything else, the details, not that great on this.

I wasn't that impressed with this one. Let's have a look at the next one, this is kind of like a Doom game and the thing I'll say about this is like it's kind of fun to play but at the same time the lighting is so so bad right, like I don't know why it's designed it like that, I've never seen AI do that before but it just doesn't look that nice when it comes to the lighting, like this is not that fun to play. Let's test the next one out, so this was pretty cool, this is like a 3D racer game and number one the gameplay is fun, number two is actually pretty interesting to look at the UI and everything else, it's built out this whole world as you can see and also like the it's smooth, it's smooth when you use it and everything else, so overall I would say this one got absolutely nailed of it. Now let's have a look at the next one, this was really really good as well, this is kind of like a Skyrim style game, you can see the pop-ups, you can see the enemies, you can see the open world, you can see the details, pretty good game, pretty interesting, fun to play, open world, I like the pop-ups on the screen as well and I like the RPG style, yeah I was pretty impressed with this.

The other thing that blows me away is like it's pretty easy to just plug this into an agentic operating system, so for example if we go over to the agentic OS over here, two ways that we're using it right now in real time, basically number one is with Quen 3.8 plugged in to Hermes and it's really really good, it's very fast, easy to use, it's designed for agented tasks anyway and also you can use it inside goal mode, so you can have it like basically coding out autonomously for hours without you using this system right here, now also if we go over to the free, sorry the Cloud Code section here, we can select Quen 3.8 as the brain of our system, we can code with it directly and we also have a workspace with everything saved inside this system as well, so when you're building out with Hermes, whether you're building out with for example Cloud Code as a harness and Quen 3.8 as a brain, really powerful system, easy to use, you can switch between the terminal and the workspace and also if a new model comes out tomorrow like Quen 3.9, no problem, we can plug it in, we even plugged in for example MuseSpark right here, so that's the way that I recommend using it and getting the most out of it, that's how we built out these tools, but overall yeah, a lot of fun and pretty good, now if we have a look at this one, again like gameplay is super nice right, I would say again like the graphics here, not the best, not the best, but when you look at the world, how smooth it is, the little houses, the open world, how far the character can run etc, like it is good, it's impressive, I've seen a lot worse, the other thing I would say is like if you use Quen 3.7, then you already know like this is a huge step up, a huge step up in terms of Quen 3.8 versus Quen 3.7, let's have a look at the next one, this one it actually beat, actually this is an interesting point, Fable 5 got beaten in many of my tests, now if you want to compare them like side by side, you can check that out on GoldieBench, we have like the side by side comparisons, but I will say for a lot of these builds, Quen 3.8 delivered something nicer than Fable 5, and I would say like Fable 5 is probably the best model in the world right now, the best singular model, but also what would be interesting is like you could use for example Hermes mixture of agents, so you could have a fusion panel of agents working together, if we go to Hermes here and then we go to mixture of agents, and you could actually use Fable 5 alongside Quen 3.8, and I think you could you could get some absolutely epic outputs if you tested and fused those models together, like that would be mind-blowing, so we'll look at the next one, so this is a video promo that we created using Quen 3.8 as well, right, so it fully created this promo video about the AR profitable room, and we just gave it the skill of remotion, and it created the animations and everything else, which was super nice, now when I actually look at this system, number one is super impressive, number two, like I think I've seen plenty of worse videos than this, the other thing I would say here is like out of all of my tests, the video one came in second, so number one that I've seen is Fable 5 for creating videos like this, but also if you want a better alternative or an alternative in general, you could use something like this, bear in mind as well, like it's not just Quen 3.8 that's dropped, but also we're looking at getting Quen 3.8.27b, which will be a dense local model you'll be able to use from next week, I think it is, it's going open source, so that means you'll be able to run like a powerful version for coding locally, I don't think it'll run so well on a Mac Studio, which is what I've got on my setup, but for everything else, you know, if you're running a DGX Spark, if you've got an RTX, that'll be really exciting, I think a lot of people are looking forward to that, one thing I will say here is like, look at this flight simulator, it's terrible, like the angle and everything else, not great at all, just doesn't look good, if we have a look at this one, this one was kind of weird, I don't really understand the point in this, apart from just being chased around, let's have a look at the next one, this was, yeah, this was pretty cool, so if we have a look at this, it's like a kind of, again, like a flight simulator that we can play, but the graphics, the details, the controls, everything is really, really good here, and also, what I will say with Quen 3.8, is that it's fairly fast, like when I was testing out with all these builds, we built all this within hours, so all this stuff was just fully automated and done, this one is pretty nice too, so it's like a dragon flight game, I like the colours that it codes with as well, they look pretty cool, here's another one, yeah, everything just seems smooth when you use it, that's one of the biggest benefits of all of this, feels smooth on the gameplay and everything else, this one was kind of crazy, but fun to play as well, now this is kind of like another 3D racer, but if you look at like the way that it moves, and the way this has so much going on when you're playing it, like it's a very interesting model, and this is the worst it's ever going to be, the other thing to say here is like, with Quen 3.8, it came out in preview earlier, and it wasn't as good as this, so it actually improved like day by day, and just got better and better to the point where it's at now, so overall, do I think it's a great model, absolutely yes, it's a really good model, it's worth testing out, it's one of the best ones that's come out over the last few weeks, it's in my opinion better than Kimi K3, it's better than GLM, it's better than pretty much all of the open source models that dropped, the actual outputs, or for coding day to day, I would still use Fable 5, I'd definitely use Quen 3.8 max over something like Opus 5 as well, which I think is like super slow, and just very difficult to get good quality outputs from, especially when it gets distracted so easily over time, but yeah, test it out for yourself, see what you think, also bear in mind, like you can just use Quen 3.8, if you go to chat.quen.ai, you can test it out for free, and also one benefit is that it can generate images and videos, which you know, Claude Code can't, so that's a big benefit as well, it's not actually using Quen 3.8 directly to do that, but it's using some of the other image models it's got built in as well. This was kind of cool, I've not actually seen this one, this is a brand new one that I've not tested at all, but with all these vibe coders building out snake games these days, it would be rude not to do one of these. This was really interesting, so this was a website, I know some people will be watching this and say like, why don't you test it with business automation, so we have tested Quen 3.8 max inside our agentic OS with Hermes already, so we use it for business automations there, but also just for like coding our website, like this website looks super nice, it's different to your normal outputs, it doesn't look like a generic Claude AI slop website, and it's actually pretty nice, pretty nicely designed, I like the animation in the background and everything else. This was a full operating system, so it's actually coded out apps and tools and games and everything else in the background, so we have paint for example here, and everything works perfectly actually, but I like the UI too.

And then yeah, just some other cool stuff as you can see, so overall great model, worth testing out, definitely one of the best on our tests. If we actually have a look at a leaderboard on Goldie Bench here, you can see the Fusion is right at the top, which is not an individual model, that's just using a panel of models. Opus 5 actually creates some really nice stuff, but it's not that good coding day-to-day, so even though it can create some pretty nice stuff, you might be looking at coding out for hours, and you have to keep it very very on track. So just for a coding experience overall, I don't recommend it, but it is capable of creating some amazing stuff.

Then we've got Hermes mixture of agents, which again makes sense because it's just combining frontier models. We have GPT 5.6 Sol and Fable 5, and then just below that is Quen 3.8, so it came in number sixth on the leaderboard so far. Impressive model, definitely worth testing. If you want to get our full agentic operating system with Hermes agent built in, you can grab that over here, so you can see our agentic operating system with Quen 3.8.

We've got a voice agent for Hermes, we have Hermes Oracle that looks at your competitors, Hermes Astros that monitors certain keywords as well, a full memory system plugged in there as well, and then also you can use the system where you can use Cloud Code, but you can switch your brain to Quen 3.8 or whatever you prefer to use. So if you want to get this whole system, you can get it inside the AI Profit Boardroom, link in the comments description, or go to the aiprofitboardroom.com. Some people say open source models like Quen, they can't compete with Frontier, they absolutely can. This was a huge step up, as you've seen in the builds that we've done earlier.

Other people say, well, you know, how do you actually get the most out of it? I would recommend probably inside Hermes, you know, that's one of the things that you can't really do with Fable 5 unless you use the API, but you could use Quen 3.8 inside Hermes and get a lot out of it, which is what we recommend and recommend you do. Inside the community, you can also post questions, get help and support. I create a video tutorial for everyone every single day, so that I answer your questions in a video tutorial, and there's always people online 24-7 as well.

Inside the classroom, you can get access to our agentic OS, and we also have a one-hour Quen 3.8 masterclass inside here as well, so you can get the video tutorial, the step-by-step guide, the zip file to install, and then also as new things drop, we add full tutorials as you can see. And then inside the calendar, you can jump on weekly coaching calls, ask questions in real time, share your screen, meet other members. Inside the map, you can meet people in your local area who are building out with AI agents like you, and this is all available inside the AI Profit Boardroom. Link in the comment section or just go to the AIProfitBoardroom.com.