Hy3 Destroys GLM 5.2? artwork

Hy3 Destroys GLM 5.2?

AI News Today | Julian Goldie Podcast

July 7, 2026

GLM 5.2 vs Tencent Hunyuan HY3 (HY3): Side-by-Side Tests + Final VerdictThe script compares GLM 5.
Speakers: Julian Goldie
**Julian Goldie** (0:00)
Glm 5.2 versus the new Tencent HY3, who wins? I'm going to show you side by side, the software we built so far, and how they perform, and who wins so far. If you're wondering what's the difference between these two models, well, HY3 just dropped, Glm 5.2 dropped about two or three weeks ago, and HY3 comes from Tencent Hunyuan, whereas for example, Glm 5.2 comes from ZAI. You can get access to both of them for free. So if you want free access to Tencent HY3, you can actually go to OpenRouter like so, and it's free, it's open sourcing, get a free API for it, and it's available until July the 21st for free on OpenRouter. If you want to use Glm 5.2 for free, you can actually get access via ZAI, and then you get access right here. As far as I'm aware, you can't get the API for free with Glm 5.2, but if you're wondering how do they perform side by side, let me show you some of the examples right here. So the first test we ran here was with Glm 5.2 and Tencent Hunyuan. All right, so this is the example from Tencent Hunyuan. And one of the things that I will say here is like it's pretty buggy when you're trying to navigate around, but it doesn't look bad. Now we did have to go back and forth with this a lot to use it. But again, you can see it's almost unplayable, like in terms of the way it's just difficult to navigate if not.
If we have a look at Glm 5.2 creation for the same model and test, let's have a look here. So this is a lot smoother, a lot easier to use. The graphics are not quite as nice, but it still can create something decent and it's much easier to navigate and play. So I would say Glm 5.2 won that round for sure. And let's check out the next one. By the way, these are both Chinese models.
So it's interesting to see that both models are open source and both have come out of China. Now, this is Glm 5.2's attempt, something called Dragon Realm, which is kind of like an open world 3D game, right? And it looks pretty cool. Like RPG game actually works really smooth, nice graphics, easy to move around. You know, that's a pretty good creation. Now let's have a look at it. And you can see here, it's not quite smooth. It doesn't feel quite as nice. And also you see how you can't really, you don't, number one, you don't really get many characters inside the game. Number two, it feels a little bit laggy and slow. So when we move around, for example, that line just bugs around a bit.
And also it's just not as enjoyable to play. Now, if you're wondering, okay, how did you test this? So we got Fable 5 to test it. And literally we were going back and forth with it for like three hours, trying to get good outputs out of HY3. It just couldn't really get there. You know, it just wasn't quite smooth. That's the thing that I found is like GLM 5.2 just first time round, it will give you better outputs than using HY3. So if you look at both of these rounds, GLM 5.2 definitely won. Let's have a look at the next one. So we've got a flight simulator here. So this is the version from GLM 5.2. Kind of interesting, like first person.
Terrain is not bad, easy to navigate, easy to use to move around, etc.
Looks pretty cool. Now let's have a look at the version from HY3, which begins pretty nicely and then kind of goes a bit wild. Like once you've flown up in the air, you can't really do anything. And the terrain just totally disappears, all the detail disappears. It's kind of a weird vibe. So I would say again, it just doesn't feel quite as detailed and fun to play as GLM 5.2. Next up, we have this driving game. So we can drive around a city, a neon city it's called, and this is a version from GLM 5.2. Pretty fun to use, still basic, but it works. We can navigate around. It's pretty smooth. There's other cars in there. It comes up with like pop-ups on the text as well. Then we've got a bit more detail here as well.
Now let's have a look at that compared to HY3's output sequence. HY3 here lacks detail, is a little bit buggy, and just kind of feels a bit more dull. I mean, like if you look at the colors and everything else, it works, but again, it just feels more basic, feels like an older generation versus GLM 5.2.

4 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/1000775850602