I Tested Qwen 3.8 Max Against Opus 5 It Was Brutal
Qwen 3.8 Max absolutely destroys Lord Opus 5.
Let me just show you. I gave both models a prompt to create a birthday app where I can track all the birthdays of my friends, family, and everyone else. The app created using Qwen 3.8 Max looks really nice. It even has this option to throw confetti, showing you live animations. It just looks so good. There’s a whole calendar for the month, and a breakdown of the next 30 days showing whose birthday is coming up. Everything works really, really nicely—you can edit people, you can delete people, and all the design and animation is just super amazing.
Then, using the exact same prompt, we received this basic, poor app using Claude. I was angry because this app took $7 to create and it almost hit my limit of Opus 5 just to build this basic thing. You can decide for yourself which you think is the better one. I know people will say they have already created a lot of apps using Opus and Claude 5 models, but we can create apps and websites using any model these days. The real test comes when you compare them side-by-side.
And guys, this model is free of cost. I am using Qwen 3.8 Max free in the Qwen Studio application, and I have not hit the limits. Meanwhile, on Claude, I’m paying money to have a really bad app.
(Just as a disclaimer: this app created by Qwen 3.8 is right now storing data inside the browser storage, but you can see that Opus 5 is really, really bad. It cannot even plan proper features, and the UI is terrible.)
The Subscription Tracker Test
Let me give you another example. I created a subscription tracking app. The one created using Qwen 3.8 looks really nice. It has a similar style to Gemini and gives a lot of things you never even asked for. I really like the design, the UI, how it is created, and the font selection. Everything looks really, really nice.
On the flip side, the app created using Claude Opus 5 is bad. The total of all the subscriptions wasn’t even correct! We couldn’t delete things, we couldn’t import things, and the OCR wasn’t working. In the Qwen app, everything works nicely. I can import using Excel, and the OCR works great. And this is from just one prompt.
I want to repeat again: I’m using Qwen 3.8 Max free of cost for hours, and I have not hit the limit. Even after paying for a pro plan for Claude, if you try to create one single app, just after one prompt your five-hour limits are reached. It’s a very, very bad experience, and I don’t know why somebody would need to use Opus 5 these days when such great open-weight models from China are available. You can just rent them for a tenth or hundredth of the price of these premium models.
Building a Hotel Website
Let me give one more example. I created a hotel website using Qwen 3.8 Max. It created the images itself, and all the text was generated. It has sliders to filter rooms according to price and amenities, and really nice minimal animation on the images. The website is mobile-responsive and very well organized. I tested it using PageSpeed Insights, and it’s loading 100 out of 100 for performance and SEO.
Where It Falls Short: Video & Images
Okay, let’s talk about video creation. It’s super basic. I gave it a thumbnail and said “animate this.” The result was very basic. Then I asked it to make a video of me punching Opus 5 and letting it explode. It completely changed to a different person punching. It didn’t work.
When I asked it to create some slides explaining how Qwen 3.8 Max is one of the best Chinese AI models, I specifically asked it to add a picture of Alibaba and Jack Ma. It didn’t work really well; it just added a random picture of an animated person. If I give the same prompt to OpenAI’s image generation models, it gets it 100% correctly with the Chinese flag and Jack Ma. Now, some people will say these are different models generating images. I understand that. But these frontier models decide what should be written on the images, and then another model generates them.
Final Thoughts
Literally, in most of the benchmarks, Qwen 3.8 Max is performing really, really great. Honestly, for me, I prefer it better than Claude. I think nowadays I prefer every model better than Opus 5. It is such a problematic model. I cannot even create a two-prompt app with a pro plan, but using Qwen Studio for free, I can create a complete app that has a better UI and better functioning.
For my everyday tasks i use GPT 5.6 sol with no doubth in the mind.
Let me know what you think about Qwen 3.8 Max in the comments! If you want to use it yourself, you can use it for free. Make sure to subscribe to nerdUP; it’s totally free, just takes two seconds, and I really appreciate it!
Try FREE Qwen 3.8 max inside QWEN studio here – https://chat.qwen.ai/. (not an affiliate link)
