GLM 4.7: New FREE Chinese AI Agent DESTROYS Google Gemini?

Today we are going to be looking at GLM 4. 7 which is a new free AI super agent from China which claims to beat the benchmarks when it comes to coding and all sorts of benchmarks. Right? So you can see GLM 4. 7. This literally just came out. It's a brand new AI agent and you can see how it performs versus GLM 4. 6, DeepC, Claude Sonet and GT 5. 1. Now additionally you can actually see how it performs versus Gemini 3 Pro. So these

are the benchmarks and we can test how it performs side by side today in a bunch of tests as you can see which we'll come on to later. So we've got for example like Kim K2, Deepseek, Gemini Pro, Claude as well and it's actually right up there. Now the amazing thing about GLM 4. 7 is that you can actually use it for free, right? So if you want to get access to this, you can get access at chatz. ai as you can see and then just make sure that you have GLM 4. 7 selected here. And I'm going to be comparing it against Gemini, which is their latest model, and just see which one actually creates the best outputs, right? Which one performs the best, etc. And we'll test out a bunch of prompts today. Here's the first coding battle

that we're going to test it with, which is build a 3D voxil art scene in a single HTML file. We'll test it on Google Gemlab Pro. Just make sure you have Canvas selected, and then we'll also do the same thing inside GLM 4. 7. You can upload files as well inside there, and you can use the search feature if you want to search as well. So, it's beginning to build out as you can see. Again, I'm not on the paid subscription or anything like this. I'm using it for free and we'll compare it against Gemini 3. 0. If you have any questions about this as we go along, feel free to ask. But China has dropped something really big here. This is a brand new model, GLM 4. 7 that claims to beat many other AIs in many of these tests, right? And today, I'm going to show you exactly how this new Chinese AI actually works, what Vibe coding means, the actual benchmark numbers, which I've already been through, and whether this could change how you bid things for your business. So, we got Zed coning out right here. It does look very similar to Night 3 to be fair. Like almost identical as you can see, very similar right there. We do have the output from Gemini 3 Pro built out much faster than GLM 4. 7 and we'll see how it performs. We actually do the same thing on Claude as well. So, I'm interested to see how this performs Claude. Let's go over here. We'll select Opus and we'll test this out as well using exactly the same prompt. And now we have all three models coding out. And this is already the output from Gemini 3. 0 Pro. Pretty nice. That works perfectly, doesn't it? We can zoom in, zoom out, etc. Opus beginning to code that out. Zed is out as well. So, here is the output from GLM 4. 7. As you can see, it looks pretty cool to be fair. I do the way you can zoom in and zoom out. However, it's not that it's not as 3D as the one from Gemini. I think the one from Gemini, I don't know. This is more kind of like a Minecraft style one, right? I would say this looks more like box art than this one to be fair. But they're both pretty cool. They both actually worked. Let's have a look at Claude Opus. So that's still building out. As you can see, we can actually initialize this as well and go through it like so. All right. So we've got the outputs from Claude Opus. Yeah, I would say GLM doesn't perform quite as well. I think the Opus and Gemini is still outperforming it. Although Yeah. I would say that's probably the best output so far. This one was pretty cool and a bit cleaner. And then GLM 4. 7 didn't create anything anywhere near as good. So let's test the next thing now.

We're going to say build a landing page for the air profit border. All right, from this description. So let's go inside GLM 4. 7. We'll plug in the information about it. Do the same inside Gemini and the same inside Opus and just see which one performs the best. Kurd says, "Good job. " Thank you very much, sir. Happy to help. If anyone has questions, feel free to ask. Like it's always more interesting when people ask questions and get involved on the live stream. So if you want to see anything today, just let me know inside the comments. Always happy to help. So this is coding out. Now we have Gemini coding now and also Opus as well. The other

cool thing about this is like whatever you're using, right? What's pretty awesome about this is like the old ways you used to ask AI for code, get broken mess, spend hours fixing it, give up, pay expensive developers, wait weeks, get something you can't change yourself or use 10 different AI tools, get confused and chased every new thing and then never master anything. Or the old way was like watch tutorial after tutorial and still can't build what you need, right? or think that you need to learn coding, feel overwhelmed and stay stuck. Those were the old ways. The new way is like you can describe what you want in plain English. AI builds you with beautiful design like you've seen today. You can use one powerful tool that thinks through problems step by step. You can access the open-source models for free with no monthly fees. And you can build landing pages, apps, and tools that actually look professional. And you can vibe code your way to working products without writing a single line of code yourself. You see these are building out. Look at that. That's pretty nice. That's going to be hard to beat. Gemini's output is nice. I'm not going to lie. Let's have a look at Claude Opus. I actually really like Claude Opus when it first came out for building. It was pretty awesome. Now we got the HTML from GLM 4. 7 being built out here. What's actually crazy about this is if you have a look at the benchmarks here, right? GLM 4. 7 for Amy Mass is scoring 95. 7, right? Whereas Gemini 3. 0 Pro is scoring 95. So on the test, this free Chinese model, this open source GLM 4. 7 is actually winning, right? Pretty amazing right there. And some people say these Chinese AR models can't compete with American ones. And this is what surprised me the most. If you look at the actual numbers from the official benchmark tests like we just did, GLM 4. 7 95. 7%. Gemini 3. 0 Pro 95%. Claude Sonnet 4. 5 87%. Right? So let's have a look at Sonet 87% over there. Right? And so GLM 4. 7 beat them both. Let's have a look at HMMT. So if you have a look here, HMMT on the benchmarks, GLM 4. 7 scored 93. 5, whereas Gemini 3. 0 Pro scored 93. 3 and Cla Sonic scored 81. 7. Right? So on the math tests, GLM 4. 7 is winning. Now to be fair, Gemini still wins on the other tests, but the idea that Chinese AI can't compete, well, that belief is gone, right? And the right belief is that competition is good because more options means better tools for all of us, right? So, let's go back to GLM 4. 7. Now, this is the page. It looks super nice. We can actually share that with people as well. So, we click on the share button here. We'll copy the link and let's go over to GLM 4. 7 here. See what it's like. Pretty nice landing page, right? I could actually use that. It's written nicely. Looks pretty cool. It's got some nice designs. It's super clean. It is a little bit basic to be honest. That's the only thing that I would improve here. Everything else looks pretty cool. If you have a look at Gemini's output, I would say that still looks better. And if we have a look at Opus, pretty similar to Gemini 3. 0. Pro, but I still think Gemini 3. 0 will probably convert better just based on my feel. I can usually tell just by looking at the page which one I think is the nicest and I think this is definitely the nicest so far. So, let's try another example here. I do think that Gemini's won on both tests so far. Let's try one

more. So, we're going to say generate a working Tetris game in one HTML file. Run this through Gemini as well as GLM and we'll test it on Opus 2. So, I think Opus and Gemini are still leading when it comes to coding. If you want a free alternative, then GLM 4. 7 is pretty good. I think that's actually replying in Chinese. Why is it replying? I don't know. But I've actually prompted in English, but it's come back in Chinese. But okay, we'll wait for that to code out. We got Gemini here coding out too and Opus 2 and we'll see what we get back in a second. Now, let's have a look

at the testing framework here. This is a framework that I've used today to test GLM 4. 7 against Gemini and Opus, right? It's the same method you can use to evaluate any AI tool. I call it the Vibe method, right? Which is visual intelligence building engine. V. Verify the claims. Don't believe the hype? Just test it yourself with real prompts like you've seen today. Then you're going to investigate practical uses. So forget the benchmarks for a minute. Can it help your business? That's the main thing, right? So could it build landing pages? Yes. Can it code out tools? Yes. All right. And then build something real. Not a toy, not a demo. Something you'd actually use like the landing page for profit boarding that you've seen today. Right? And then from there, just evaluate the output. Does it look professional? Does it work? Could you show it to a customer? That's what you want to test today. And we've tested a bunch of things that look really cool.

If you want the full 30-day learning path, I've got it inside the notes from today. That's inside the AI success lab link in the comments description. Completely free community with 40,100 members. And inside the classroom here, you can learn exactly how to use this stuff, right? So, if you go to December 2025 notes, you can see all the latest updates. And you've got the notes on GLM 4. 7 over here. So, we've got Neon Tetris. That's actually not showing anything. So, that's totally failed on this move. Gemini is working perfectly, I think. Oh, no, it's not. Look at that. It's just disappearing. You can't It's just dropping down the bottom. All right, let's have a look at Opus. see if hopefully this works, but you never know. Oh, this is pretty cool. There we go. I think Opus has won this one by a

long way. You've seen the results today. Would I say GLM is going to replace Gemini or any of the other tools right now? Absolutely not. Like, it's cool and it works, but I think Opus and Gemini is still superior. Even if you're on the free plans for them, you're still going to get a lot more value out of them, right? And you can see this working prototype that works perfectly, whereas GLM 4. 7 just totally failed on this. All right, thanks very much for watching. If you want to get access to all my best prompts, tips, workflows, tutorials, processes, etc. Feel free to get that inside the AI profit boardroom link in the comments description. It's an amazing community where I personally connect with all the members. We've got 1,800 builders inside there who are serious about using AI for their business. You can see inside the community here. You can post questions, get help, get support, comment on other people, lurk in the shadows if you need to, and you can just connect with loads of cool people. You can also get this map where if you're in a local area in the world, let's say you you're in some random place like Chile, right? You can actually meet people in your local area just by going on the map and finding out who's in your local area. As you can see, right? So, for example, let's say you are in the UK, right? In the UK, you can see all these members that are all over the UK and then you can DM them and say, "Hey, you want to meet up? Should we get to know each other? " Or you can just jump on a Zoom call with them online like you see here. So, Matt is based in Manchester. We've got someone in as well. This is Stefan, right? And so you can meet people on the map here and it's still private because the radius is like 10 miles so you don't really know where people are. They're within the local area and then you can DM them, jump on a Zoom call, maybe meet up in real life and it's super easy to connect with people, right? Also inside the classroom you get all of my best automation. So if you've never used AI before, you don't know where to start, just follow the sixe road map here that takes you from a complete beginner to learning and mastering AI. Also inside the classroom here, you get access to my latest SAP updates which are dated so you know exactly which ones came out recently. You get all my best playbooks for implementing this stuff inside your business. So for example, if you want to know how to automate your emails or automate Twitter, automate newsletters or shorts or avatar videos, then you can get all of these trainings inside the AI profit boardroom as well. All right. And if you need even more help, you can jump on live coaching calls with Dr. Jeff, a trained PhD master in AI. And you can jump on the calls with him. live coaching calls, get help, get support in real time, plus meet the whole community inside there. Top of that, we have a full agency course that shows you how to get more clients for your agency. Or if you want to watch the coaching call recordings, you can watch them back here. If you want to get my best, you can go here. YouTube AI road map. This is a full six week road map that shows you how I grew my YouTube channel and yeah, just all my best trainings are inside the classroom here. So, feel free to get that link in the comments description. Appreciate you all watching and I'll see you on the next one. Cheers. Bye-bye everyone.