China's GLM 5.2 is now FREE!

**Julian Goldie** (0:00)
Today, I'm going to show you how to use Glm 5.2 for free. Now, bear in mind, Glm 5.2 is basically frontier level at this point. It's on many tasks, it actually outperforms Opus 4.8. I'll show you some examples in a minute, and you can get free access via Zed AI. It's just been added recently. Now, there will be token limits on that, but if you want to use it for free, test it out, see if it's right for you, et cetera, then you can go directly into Zed.AI and then start using it, as you can see right here. Just make sure that you select the model, it's Glm 5.2. There's two different options, so you can use the agent over here, and then you can use the chat.
So if we test this out, for example, and we'll say, you know, are you working? And then what you could do inside here as well is you can add web search, and you can also like attach images and that sort of thing. And then you can switch between high and max. Obviously like max would use more tokens, and also deep think would use more tokens than non deep think, right? You can switch between thinking and non thinking. And this is an open source model, but you can also get access to it for free via this section, as you can see. And if you don't have deep think on this, it's super fast to reply. If you go to the agent section as well, we can start building with this too.
So we could say, okay, build out an SEO tool over here, hit enter. It's going to be using Glm 5.2 to start building out, and that will just work in the background. Probably take about 10, 15 minutes to work in the background, I think, when you're using the agent section here. Now you might say, okay, why would you use Glm 5.2? Well, it's a Chinese AI tool. It's open source. You could use it for free, and it's right up there with the most powerful model in the world. Out of everything that I've tested, and I have tested stuff relentlessly every single day, out of the tests I've seen so far, Opus 4.8 is winning, but Glm 5.2 is still up there. And if we actually look at the tests here, you can see how it performs. It's a million-token context window. It's open source. I tested it on 21 different tasks. It scored 8.23 out of 10 on my own Goldie Bench benchmarks. You can see it was just released last week. And also it won quite a few medals on my tests as well. So if you're wondering, okay, like how does this perform, et cetera? What is it? So Glm 5.2 is Zed AI Frontier model for million-token context window, released on June 14th. The tagline is never forgets agent, million-token context window, open weights, which means it's available. You can get the weights on Hugging Face for self-hosting or it's runnable for free on Zed AI as well. And so I actually use it inside our agent operating system, but you can use it for free over here. Now, if you're wondering, okay, what did you build with it? Let me show you an example. So you see this is quite slow to use by the way on Zed AI when you use it for free. But if we go to Glm 5.2 over here, we can see everything that we've created with it. So you can see, for example, we created like, you know, like full 3D games with this stuff, which was pretty fun. As you can see right here, we can also plug it into our Hermes agents as well using the CLI.
And it creates some pretty awesome stuff like, this is pretty fun to build, pretty cool to create. Here's another example. So this is called Neon Drift. It's kind of like a racing car game, as you can see.
Fully playable, actually works quite fun to play. And we build that again. Now, if we have a look over here, we've got this full open world game as well, where we can just navigate this world, fully created in one single prompt, just by asking GLM 5.2 to build it. Here's another example. Here's another example. We even, for example, created like a landing page, a voxel game runner as well. And the power of this stuff now is like, you know, at this point, I honestly think, oh, and if you look at the benchmarks as well, at this point, it has overtaken GPT 5.5. It gives better results than GPT 5.5, and it's way ahead of something like Gemini 3.2 Pro. Now, if we actually have a look as well here, it created a full operating system. So you can see that we have an operating system with apps, like it's got a notes app, it has a terminal, we have the music player inside here, and do you know what else blows me away is, this is music we created inside our agent OS recently, inside the music agent, and we can just play it through this operating system that we created with Glm 5.2. You can see here, it's fully works. Like this is a full operating system, we've got paint built in, we've got the notes section here. If we type something, then close it, then open it up. Again, it actually saves what we created.