Cancel Your Claude & ChatGPT Subscription! Ox-Alpha Is HERE! (GLM-5.3 Flash)

A free AI model just showed up and people are already putting it on the same level as Chachi BT's Soul 5.6 and even Claude's Fable 5. Now, this one's called Ox Alpha. And the big question is whether it can actually deliver similar results without costing you anything or whether the hype has gone too far. Now, we're switching things up. Most videos show you how to set it up through a terminal, which can get pretty complicated pretty fast, but I'm doing things a little bit differently. So, really, anyone can open it up and follow along. So, we're going to give it a real world task and compare its output, its speed, and the final results against the AI tools that you already know and use. And by the end, you'll see exactly where OX Alpha shines, where it falls short, and whether this free model can genuinely build something useful or if it's all just pure AI hype. Now, for this, we're going to go ahead and keep everything extremely simple, simpler than anybody else has showed you. So, what we're going to be doing is we're going to be logging in directly to Z.AI where OX Alpha or GLM 5.3 Flash is being hosted. That way, we don't have to set up any complicated terminals. We don't have to do any complicated coding. We're just going to get straight into the action and we're going to compare what GLM 5.3 Flash gives us against ChatGBT's 5.6 Soul and Claude's Opus 5. So, we're just going to go ahead and get started and we're going to create a simple page for let's say a photographer. We're going to do a landing page for a photographer. So, the first thing that you need to do is run over to Z.ai and simply just log in. Once you log in, your default model up here is going to be 5.3 Flash. Now, we do have another more intricate model or more advanced model, which would be the 5.3, but we're going to stick to 5.3 Flash because this is the one that's using the less amount of tokens, the less amount of credits, and you can use practically for free, and it's supposed to be side by side in terms of output with Opus 5 and Soul 5.6. So, we're going to be the judge of that in just a few minutes once everything is done. Now, before we do get started with our prompting, let's see a few of the different things that I can do. So, it can create landing pages. It can do teaching material, 3D modeling, miniames, and personal blogs. And this is all with a lightweight 5.3 flash version that we're not going to be paying anything for. So, let's just do this as a quick preview. Let's see what it made or what it can make. So, here it gives the example prompt. And here it actually comes up with the website. So, this looks pretty good. Now, this does look I don't want to say it looks generic because it does look pretty good, especially with the background image and the way that it's transitioning. does have a couple of errors, but overall it actually looks fairly well. It doesn't look like, I guess, AI slop, right? The typography on here is pretty nice. It's not that copy and paste font that pretty much everybody or every large language model is using. And it's got some decent little animations. It has this scrolling over and it has the different transitions as you're going through the page itself. And you have your little book free trial. That's pretty cool. So, when you hover over it, it actually has an animation. Okay, so this looks pretty good, especially for a model that's supposed to pretty much rival flagship models like Chachi BT and Claude's Fable, Opus, and Soul. Okay, so what I'm going to do is I'm going to go ahead and paste in a prompt that I have here. Now, this is a pretty big and detailed prompt. You can make it a lot simpler, but I want to see how robust it can make things and really how good it compares against the other two models. Now, if you do want this prompt, just wait a little bit longer. I'm going to give you a master prompt cheat sheet with all the different links that I'm showing you in this video along with this prompt and anything else that I'll be using. So, stick around for that. Now, what I'm going to be doing here is I'm just going to go ahead and paste it in. Again, this is a very long and robust prompt. If you don't want to wait for me to send you the link to the cheat sheet, then something else that you can do is either go to Chad GBT or go to Claude and simply tell it what you want. Tell it the type of website that you're making and tell it that you need a prompt that you can paste on whatever app you're using or whatever AI that you're using. That way you can get a strong prompt, something that's a lot more detailed than what you would typically come up with. So that way you can really start making these on your own. So I'm going to go ahead and put it on here. And just so you know, this is currently set to max and deep thinking. So you know what we're actually going to do is we're going to set all three different ones together. Now, we're not going to have too much space, but we are going to be able to see all of them working. All right, perfect. So, I have it pasted here. I'm going to go ahead and paste it in claude. And I'm going to go ahead and paste it in chat GBT. So, let's just go ahead and do enter, enter, and enter. And now, let's just give it a couple minutes and see what it comes up with or what they all come up with. Okay, so it looks like we have our first winner, which is Claude. Claude went ahead and finished up the design plan, and it's currently laying everything out for us. So, let's just go ahead and give it a few minutes. Catch unfortunately, had an error. So, I'm just going to go ahead and try it out on a different browser really fast and just let it do its thing. Now, if we look at Claude, it's currently writing all of the code for the website that we asked it to make. So, it looks like right now the winner is going to be Claude in terms of speed and we are using Opus 5 High. We're not using Fable 5 because realistically Fable is extremely expensive and not all of us are going to be using Fable just for a simple website. Even then, Opus 5 is still kind of overkill for something like this, but it should do a way better job and a lot faster job than any of the other models. So, Opus 5 is good to get started. Fable 5, we really don't need that much power just yet. All right, so I'm going to let them all continue working. Let's just give it a few more minutes. >> A few moments later. >> Okay, so it's probably been I want to say 15 20 minutes and Claude finished first. Now, Claude did just give me code at the beginning. So, I asked it to make it into a file that I can actually open into an HTML file, and it went ahead and did that. Now, unfortunately, Chad GBT has been down for the majority of the day. I just found that out, but it seems to still be working now when I ran it locally on my desktop app, but don't worry, it's still using the exact same parameters. So, it's using 5.6 on extra high. Now, Chadbt is currently working. GLM 5.3 Flash is also still thinking. It's been quite a bit. I haven't seen it do anything else. Let's just keep giving it a little bit more, but let's see what Claude actually came up with. So, this is what Claude generated. Let's go ahead and refresh it. And this is supposed to be a portfolio for a photographer. And it was supposed to receive or it was supposed to get all the images from a website called Pixels. Now, for those of you who don't know, Pixels is a completely free website that gives you stock images that you can use. Now, looking through this, it's actually really nice. So the text itself is kind of the generic text that most AI websites get. But besides that, everything else actually looks pretty clear. So here we have a couple of examples. We have here a drop down. Oh, that's pretty interesting. On drop down a where you can click and it'll take you to that section of the website. So that's pretty nice. The animation as it scrolls down is really good, too. But let's see. So we have a working sheet, 14 frames, 4 years. I don't know about that, but I mean that's fine. And we can go ahead and change anything if we choose to. The face over here is cut off, but everything else looks fairly good. It does need a little bit of extra love in terms of, you know, the layout and how everything is kind of positioned. So, here you can see that she's cut off in the middle. Here, her head is cut off. And this one's all right. Scroll through. We have different sizes. I don't think that this should be completely different sizes on the same section. There should be something down here just filling in that extra space, even if it's just a little bit of text or something like that. There's also a lot of images in just this one clutter. So, it's not horrible, but it could be formed a little bit better. Now, if we switch over to the different tabs. So, if we do portraits, it only shows the portraits. We have one here that doesn't seem to be loading. And then if we switch it over to weddings, it brings up the weddings. These are definitely not weddings at all. It's just this one. And then travel. So, these look all right. Well, not these. This one's really the only one that looks like it would be travel. But that's okay because again, Cloud is kind of just assuming all these different things. It takes the images and it analyzes them and it thinks that this is where they could go. Or when it looks up certain images or certain search phrases on Pixels and Pixels gives it these results, that's what it's going to assume that they are. So, I see where it's coming from. I see why it made this decision. Of course, AI is not the end all be all. It's always going to need some sort of input from us. And something like this is definitely where we would come in and tell it, look, this doesn't work. These images are not relevant. These are different sizes. And then we start to fix everything from there. But if we keep scrolling, we have our understudies here. So it has a bit of information on the photographer themselves. So it looks like our madeup photographer is from Mexico City. And here's a little bit of about me. I started this in a corridor at thero de laudad waiting for a magazine portrait that kept getting pushed back. That's pretty interesting the way that it comes up with a bio. And then if we scroll through, we have a few more images here. We can't really click on anything. Now, let's scroll down a little bit. This part right here is actually very, very impressive. So, what I do to a file and what I don't. People ask what happens after the shutter. Not much. I set the black point, lift the face a third of a stop, and leave the room the color it actually was. Drag the handle to see the flat scan on the left and the finished print on the right. So, here, this is really cool. It actually color graded this image, and it did really well. So, it made it from something that's very low saturation, not a lot of color to something that actually pops. So, this is actually a very nice touch. Now, scrolling down again, we have more information on the photographer themselves. Her name is Marasol, and she's been doing this since 2011. Here we have Marasol in the Huarez room. Clearly, that is not the right image, but again, this is all just completely made up. We would have to make a few adjustments on our end. We're just trying to see how this actually just overall looks and feels. Now, we keep going down. This is pretty interesting. We have a pricing list. So Claude came up with different pricing for different types of shoots. So the editorial portrait sitting is $2,200 for half a day. Weddings available light. So I guess bringing in a light source 6,800. By the way, I used to be a photographer, so I kind of know some of the different things that it's going for here. And documentary and travel commissions 1,900 per day. And prints 180 each. Then we have what people have said afterwards. So we have a bit of a review. This right here does have a lot of empty space. This would be something that needs to be filled in either with let's say example images or simply just take the reviews and line them up side by side. But keeping going, we have tell me a contact us form. So tell me about it. And we have our different links down here. Now, unfortunately, none of these work, but again, that is to be expected. And of course, this file is not online. This is an actual website. This is something that's running locally on our machine. And if we do want to put this up online, we're going to have to find a hosting software somewhere else. Now, let's see where the other two are at. So, it looks like ChachiBT is still doing its thing. Again, perfectly fine. It errored out the first time on the web version, so I had to switch it over. But, let's see what GLM Flash is coming up with. Now, it doesn't look like it's fully done yet. It is finishing up. It looks like it's done quite a bit of work. This actually looks a lot longer and a lot more detailed than the one that Claude had sent me. So, let's go ahead and give it a few more minutes. Let it finish up and we will be right back. Now, I did promise you how you can gain access to the prompting cheat sheet. And if you want access to that, just drop a comment down below letting me know if GLM5.3 is something that you're actually going to use and let me know your thoughts on it along with the # oxalpha. And when I see that you went ahead and did that, I'll reply back with a link to the sheet. All right, let's continue. Okay, so chatbt finished up and this is what it gave us. And I actually think that this is much better than what Claude gave us. These are all going to be very similar, but they are going to have a couple of structural changes depending on how creative the AI models got. Now, for Chad GBT's Soul 5.6, this is what it gave us. And so far, the actual layout looks really good. Let's do a quick refresh. You can see everything kind of just fades in. Let's check out the text over here. So, stay for what happens next. I do like this font. It's not very, I guess, AIish. It doesn't look like those flyers that you see generated everywhere. Now we have editorial portraits, road stories and weddings made in the light that was already there. Now I photographed the half second after the pose, the laugh that breaks it, the why they're moving in, the hand reaching across the table. That's pretty interesting. I kind of like that. See the recent edit. See if this works. All right, so it brings us down. Let's see everything up here. Our about Okay, so everything takes you to different sections of the page itself. So the transitions looked really nice. Let's do a refresh one more time so we can see all of the transitions as we're going down. Now, if we start to scroll down, we have a working contact sheet. Portraits begin with conversation. Travel work begins by walking. Wedding asks for both attention without interruption. I'm not a huge fan of the copy. I feel like it's too short and it's trying to do too much with too little. I think something that would just be a bit simpler in terms of the text would be better. But overall, it's got a nice feel to it. Now, they all used the exact same images. And then if we scroll through, we have Andre between takes. Okay, that's pretty cool. So, it actually knew what this was. It was a person and it labeled it as Andre between different takes. Okay, that looks good. Riverbend before the trail opened. Also really good. This is something that Claude didn't do. It didn't actually analyze the images to see what was in the images. Now, if we keep scrolling, we have Mina waiting by the window. Okay, so this was good, but she's not by a window. She's actually outside. And then we have the sky kept moving after midnight. Loen. Now, again, everything's looking good so far. or we can click into these and they zoom in or they expand. We have a bit more information here. So this is in Norway in 2024 and we can click on next to see the other images. So that's pretty cool. It actually opens up a gallery when you click into them. Now the only thing that I'm not feeling right now is the actual size of these images. They're really big. But if you scroll down, we have a few more. They're not properly aligned. They don't always have to be aligned. Sometimes a little bit of misalignment. Something's a little bit higher. Some a little bit bigger. a little bit smaller. It does look good, but the way that it's doing it here, it's not really working in my opinion. But let's check out the rest of the text here. So, Camille at the end of the workday, the dress, and Dr. Nia Brooks after the interview. This is really good. ChachiT is actually analyzing all of these images, and it's making a real guess as to what they could be or it's giving them the right type of title. Now, let's continue. So, we have more images. You see, like right here, this one's huge. just could have been a lot smaller and there could have been some extra space here. This really wouldn't be very mobile friendly. And on this one, there's just too much empty space. The image is way too big. Now, we have the weather between towns. I really like how this is transitioning. It keeps the text over here while it just simply moves the images. So, this looks good. Again, little bit of extra space, but that's fine. Now, this is Laya. I'm Laya. I notice the edges. So, it's supposed to be our photographer. Now, this was the same image that was used up here in the breakroom, and her name here was Sora. So, it doesn't really make sense there. It doesn't coincide, but fine, I get it. Again, we're using stock images. We're not giving it an actual photographers's profile or their real work. So, it gives us here a about us for her or about me. Then, we have the clear plans, room to wander. We have their pricing. So, the editorial portraits, the wedding stories, small gatherings, and we have the different pricing over here. Scroll down. We have what it felt like. Cool. So, this is a review section. It looks good. I just feel like again, it's just the text is way too big. And you don't really know that this is a review section. You really actually have to read everything through here to know what it is. And then we have the contact us form. So, so far it's actually not a bad layout. The layout itself to me is a lot cleaner. It's probably a little bit smoother than what we saw with Claude. But now, GLM 5.3 just finished and let's see what that came up with. Okay, so right off the bat, everything kind of just faded in. Looks pretty good. I like how this is a little bit crooked. It gives it a little bit of character. The text itself, the font doesn't look too bad. It kind of reminds me more of the Chachi BT one. Same thing for the text down here. So, I don't carry lights. I wait for the window. This is actually pretty cool. So, this makes it a lot clearer that this photographer doesn't use flash. They use natural light, which is a good way to be able to put it. Now, we can go ahead and see the work or book a session. If we book a session, it takes us all the way to the bottom. Okay, let's do another quick refresh and let's see how everything scrolls through. So, selected work, three bodies of work live here and shows all of the different images that they did again with no flash. Now, one thing that I'm kind of iffy about actually that's really nice when you switch through the tabs, the way that they move. But one thing I don't like or really don't understand is this little red circle. Maybe it's supposed to highlight our images. I'm not entirely sure. But this doesn't really work for me. This doesn't do anything. I would take that off completely. Now, if you look at the text, right, so this one is Tomas before his shift. Okay, cool. This one is first dance window. The context here isn't really working. There's no dancing and there's no windows. There is this arc over here, but it doesn't really go with what GLM is trying to say here. Then we have the last swimmers. There's nobody really swimming here, but at least it is a river. We have confetti borrowed garden steam and neon square. Donna Alice in the good chair. Okay, so the text itself is not really relevant to the image itself. So what Ox Alpha chose or what it analyzed these images to be, it was wrong and pretty much on all of them. So that's before the fleet off the wall at 5. That's not really working. Now if we scroll through, we have the sitting room. So we have information on our photographer. Yeah. So since 2018, I've been photographing people in the rooms they actually live in. Okay. The cost of twins. There's no twins. There's only one person. Here we have a broken image with a little animation. Same thing here. And then here we have the before and after. So this is also another really good thing. This is something that Chacheti didn't do. It didn't give us the before and after. And I really like this because it shows how somebody can take a basic portrait or a basic picture and just make it a lot more vivid, a lot more live. So this little bar here I think was missing from ChachiBT. I think this is absolutely necessary. Once I saw it, I thought this is a great idea. But then if we keep going through, we have some more information on our photographer, their rates like in the other ones. Now, interestingly enough, this is actually in euros. So, I'm not entirely sure why GLM decided to go with euros rather than dollars, but it's fine. It is what it is. Okay, this is actually nice. So, this was labeled now client words and it gave us the reviews. So, this actually makes it a lot more obvious that these are reviews. Then, if we scroll through, we have the tell me about the light. This is our contact us with the contact us information. So overall, I think everything came out pretty good on this part on GLM 5.3. Now, if I had to score all three of them, which ones were the best, which one was the worst? This is my final verdict. So, in terms of the total layout, I do like the color scheme that Claude chose, and the layout is pretty similar to all of the other ones. Now, Claude did fill up the images a little bit better than some of the other ones did. It didn't make them huge and overbearing. It did miss a few spaces in terms of text. that has a lot of empty spaces, but the images themselves are correctly proportioned. Everything is the right size and it's not overbearing. Now, Chachi BT chose a very nice color scheme and the overall look and feel of the website is good. It actually analyzed all of the images and it told us what they are all about. It did have a couple of errors here and there, just like this one saying that she's waiting by the window when she's actually outside. But everything else is pretty on point like this one like we mentioned earlier Camille at the end of the workday over in Queens. ChachiBT did however miss one very important part to the whole website and that was the before and after with the color grading. I think that would have been a very nice touch to this. Another thing about Chacht's output is that the images are huge. Like they're way too big. Every single image on here needs to be resized and everything kind of just needs to be compacted a little bit better. But overall, the look and feel of the website, it looks pretty good. The font and the text is all pretty on point. It doesn't look too generic, and it doesn't look very AIish. Now, keep in mind, these two models that I just showed you, we're paying for both of these models. Now, if we open up the one that Ox Alpha made for us, this one is pretty similar in terms of the color scheme to Chad GBT. And as far as the layout, it got everything kind of similar to Claude. So, I feel like this is a nice little mix of both of them. And it's very impressive because this is the completely free version. It did take a little bit longer than the other two, but it did a pretty good job for something that's totally free that you can just sign up for right now and get done. Yes, there are a few things that we need to fix. A few things that really don't make sense. Like I mentioned earlier, these little red circles, they don't really work. They shouldn't be there. I don't understand it. And the alt text or the text for each one of these images doesn't match. But again, these are things that we can easily fix. And I didn't notice this up until now, but as I'm scrolling through everything, it's all in just this one consolidated section. So it's not everywhere on the page and it's not overwhelming the users or the visitors. So overall I think for a completely free version it doesn't 100% compare against chat GBT or Claude but Ox Alpha did really really well. Now is Ox Alpha actually as good as everyone is hyping it up to be in my opinion for what it was able to do for completely free? Yeah, it's pretty good. Again it's not up to the level where Claude Opus 5 is or Chacht 5.6 Soul. It's not at that level. But again, we're going off of a completely free version here. So, while it can make pretty nice and intricate websites like the one that I just did right now, like a landing page, I'm not entirely sure how well it could handle creating an entire storefront, like let's say making a backend to it, attaching a payment processor, or coming up with a game. Now, if you do want to see GLM come up with any of these different ones and how it compares with any of the other models like I just did, make sure you hit that subscribe button, make sure you like this video, and turn on the bell notification so you get informed the moment that we upload a video. We're going to be covering a lot more AI topics and how you can start making money with that. Now, as far as the usage for it, so as far as using the actual model, it's just the exact same thing as when you're using Claude or Chat GBT. The way that we used it in this case is by simply talking to it as a large language model. It does have the ability though to connect to Claude Code through the terminal. Now, if you do want to learn more about that one and see how that works, let me know down in the comments below. I see a lot of people actually covering it that way. I decided to make it a lot simpler so that way anybody can get started with it the moment that they watch this video. Now, as far as the compatibility or how you can connect it with other apps, this is something that we weren't able to test and I don't see the option to be able to connect it to anything else to have any connectors attached to it. So the way that we use chat GPT with Shopify or Claude with AutoDS the way that we connect it all together through either MCP connections or different plugins, I don't see the option for that just yet on GLM 5.3. Maybe I need to dig around a little bit more. But again, as it stands, it's a completely free model. So for what you're using it for and for what essentially you're not paying for, it's definitely 100% worth trying. So remember, let me know down in the comments below if you actually want to try out GLM 5.3. Is it something that you're going to use? Let me know that along with the hashtag oxalpha. And once I see that you went ahead and did that, I'll send you over the prompt sheet with the example prompt that I used earlier to come up with the photographers's website. Now, if you do want to learn how you can use any AI or specifically Claude to build a $10,000 website, make sure you check out this video right