China's Qwen 3.8 Max DESTROYS GPT 5.6? artwork

China's Qwen 3.8 Max DESTROYS GPT 5.6?

AI News Today | Julian Goldie Podcast

July 20, 2026

Qwen 3.8 Max Preview: Alibaba’s 2.4T-Parameter Open Model Claims #2 Behind “Fable 5”The script covers Alibaba’s announcement of Qwen 3.8 Max Preview, described as a 2.
Speakers: Julian Goldie
**Julian Goldie** (0:00)
Qwen 3.8, huge announcement from them today. I've never really taken their models that seriously. Obviously, this is Alibaba's team, so they are pretty good. But at the same time, I've never seen a model that claims to be this high level from the start. So what we have here is Qwen 3.8 Max Preview. It's already made its debut on Alibaba's Token Plan, Coder and Coder Work. So you can be one of the first to try it out. I've tested it out on the website, and I can't seem to get access to it just yet, and I can't find the API on OpenRouter just yet. I know it is available in China, but this is a 2.4 trillion parameter model, and here's the most important thing, my friends. It is second only to Fable 5 So they've said, we believe it's one of the most powerful models available today, compatible to leading frontier models, second only to Fable 5 That is wild. If you're not familiar with what's happened this week, Kimi K3 from China came out, which is another open source model. Remember, this is open source as well. This is going to open way. Usually, you can get free access to their models available at qwen.com as well. And their Qwen 3.7 model was actually free for a little bit last time it got released. So I'm excited to try this.
And what we're looking at here is a serious situation where basically China has leapfrogged ChatGPT in just the last week. So for example, what we're seeing here is that Kimi K3 came out this week. And that was already right up there, right above something like GPT 5.6. So it's pretty much on par, if not slightly above on front end coding when it comes to Fable 5 But what we've also seen here is that Kimi is not even mentioned inside the announcement from Qwen. They have not even said that it's like, okay, second only to Kimi K3. They've said that the only model available today that is a second only to is Fable 5 So this is a really big announcement. I've never seen Qwen go that bold. They are firing shots, my friend. And we're looking at a really interesting situation here, where you can get Qwen 3.5 Max Preview, sorry, Qwen 3.8 Max Preview. They do have available plans as well on their websites. You can go with Qwen Cloud. And this is one of the most important parts to understand. Joe actually posted about this, which is that if you really look at what is going on here, open source models from China are beginning to overtake the models from the US. We are looking at a world where China's models are becoming more and more frontier, even though they're open source. And also bear in mind, we've still got GLM to drop their next update. I'm also looking at DeepSeek GA4 Pro that should be dropping very shortly as well. So China is closing the gaps with US frontier models. It's getting closer and closer. If you haven't already seen what Kimi K3 was capable of, because for me personally, I don't really listen to the benchmarks so much. So when Qwen 3.8 Max does drop, I'll be excited to check it out. But I do test everything out personally.
I tested out 50 different builds with Kimi K3 and it was ridiculously good. You can check it out at Goldie Bench. But the thing that I saw is like it's unbelievably good at pretty much everything that it does.
The quality of this stuff is super nice. The ambience and the vibe of everything that you coded out is pretty good. And if Qwen 3.8 Max Preview is even better, I don't know what's coming next because these models are not just like competing with the US, they're not just overtaking the US in some situations, but also they're improving more rapidly.
The improvement cycle between models from China is like spinning faster and faster if that makes sense.
And also they don't seem to be like delaying the models in preview mode or like beta testing before they come out. They just drop out to the public straight away, which is pretty much what the public want at this point. And we also have more waves coming. So we've talked about Kimi, and we've talked about, for example, Qwen, but also you've got DeepSeek and you have Minimax. Minimax have been quiet for a while, so I'd expect the competition to rise and the standard to rise even higher there. But the main point here is like open source is getting bigger, better, improving faster, no delays in the previews or the releases.

2 more minutes of transcript below

Feed this to your agent

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/1000777635673