NEW GPT 5? Quasar Alpha is INSANE (FREE API!)

So today we're going to be looking at a new stealth model, very mysterious, called Quaza Alpha. And nobody is sure what it is, but some people are saying it's the new Gemini 3. 0 Pro. Other people are saying this is the new GPT5. We'll be testing out today. So you can actually test this model from Open Routter inside Openroo. ai. Let me pull this up for you

and we'll show you exactly how to get access to it. All right. And you can see it's actually right at the top here trending directly on open router. So if you go to openrooout. ai and then you go to quaza alpha. This is a cloaked model. Nobody is sure what it is. Some people are speculating it is for example GP5 or Gemini 3. 0. We're going to go inside the chat here and we'll compare it as well. So we can actually do a comparison versus other models. You can see it's got a 1 million token context length which is absolutely insane. and it is completely free to use right now. Now, from my test so far, it is very fast to use and you can also code with it directly inside of Root Code and client. It is absolutely insane what you can do with this. All right, so for example, we

can actually do some tests here. So, let's say for example, we wanted to build a self-playing snake game. We'll say game inside the chat here. And in a sec, you'll see it respond. Again this is completely free to use and it's now creating the code directly. We can also say okay create that in HTML and it will just go off and do its magic. All right the other thing you can do here is you can go inside open router go to your keys create an API key and then you can start using this inside other models. Right. So, for example, if we go inside

Visual Studio Code here, then we're going to go to settings. Then, we're going to select open router and we'll switch from llama 4 over to, for example, quazar. All right. So, open router/quaza alpha. We'll hit save, hit done, and then we can do its magic. Correct. Now, this is a model that we built, a self-playing snake game we built with the latest release from Llama 4. And you can see it is pretty trash. All right, it doesn't work at all. Now, if we try the same thing with, for example, Qua Alpha. So, we're going to say create a self-playing snake game. And I'm just going to switch to normal code mode here just to make it easier and faster. It's going to start coding with that free API key. All right. Super simple and easy to use. Now, if we go back to Open Router, we've got the response in HTML over here. as you can see. So, we can copy that and then we'll test out on liveweed. And that is working perfectly. Look at that. Boom. Even a nice little UI. So, you can use it directly inside the chat here. Super fast. And you can also use it inside Visual Studio Code. I would recommend using it in combination with something like R code for coding. And then you can build out free websites and that sort of thing. But it does seem super powerful. You've seen the output here of Qua Alpha versus the latest model from Llama 4 that literally just came out a few hours ago and this is much better. All right, so pretty simple and easy to use. Just going to go back here. Make sure that we are using quaz alpha seems to be some sort of error where sometimes if you hit save it doesn't work right. So we'll switch that back now and we can

also for example like build website. So if I say okay build a modern sleek landing page for my SEO agency, we can build out websites for free as well using Quaza Alpha. All right. So you can see here the API cost is zero. It will just start doing its magic and building stuff out. Now if we actually go inside the chat models here and we say, okay, what model are you? It says I'm based on OpenAI's GPT4 architecture. Which makes it everyone think like this must be a new model from OpenAI. Right. So we'll say okay what's the name of your model? Are you GPT5? And it says no. I'm based on GPT4 not GPT5. It has knowledge up to October 2023. So it's still part of the GPT4 family. Okay. So if you see any speculation saying it's GPT5 at least you know now right

now if we go on to YouTube here a lot of people are raving about this model right so you can see here for example some people saying how good it is other people saying this is like a cloaked new GPT5 etc. Either way, it is pretty powerful and it's been good so far. So, let's see what it comes back with on R code here when it comes to coding this website out. I also did some deep research on Qua Alpha and it was saying it's a mysterious large language model recently released by OpenRutter sparking speculation about its origins and capabilities. Key features 1 million token context window. Bear in mind, L of apparently has a 10 million context window. So even though this is large, it's not the largest stealth release. So it's released quietly as a stealth model for free and performance. So it's showing really strong benchmarks like you see right here. You can actually see a couple of the benchmarks in terms of how it's performing. This is an awesome article by Julang and you can see how it's performing on the ADA polyglot coding benchmark. So it's competitive with 03 mini medium, deepseat version 3 and Sonic 3. 6. is also very fast. So Qua Alpha is here and it's pretty much holding its own against Claude 03 Mini and Deep Seek, but it's being outperformed by a long way versus Gemini 2. 5 Pro, which honestly I think is by far the best model that I've seen so far. Also, other people are saying there's some technical clues here. So the use of chat cmpl prefix in API responses and a tool called ID format similar to openis supports those open AI as well right so it must be a model by open because inside the chat here it's saying it's based on open's GP4 architecture and then also inside the prefix it's also being used there as well let's test out for writing as well so I'm just going to go my list of tests We'll grab that and we'll see how it performs for creating content. What I'm also going to see is does the web search method work. All right. So, even though it's only updated to October 2023, if it works with web search, then that's an easy way to like just make the model much more powerful. Also going to do this on a new room simply because we don't want to ruin the context before. And you can see it's coding that out right here. And now we have the page ready to go. So, it's created that landing page based on SEO agency. It's not as good as other models I've seen, for example, like 3. 7 Sonet, but it's not bad for a free API, right? And you can easily, for example, automate and build and code out tools, apps, games, etc. inside Visual Studio Code. Completely free API. You could also use this, for example, inside NA10, I believe, as well. So, let's test it out.

We'll go over to NA10. See if this works. So, if you want to build out your own AI agents, we'll see if we can connect this to Qua Alpha. So, what we're going to do is select router. We'll grab a new API key and then we'll plug that in here. Perfect. All right. Can we get Qua Alpha here? Yeah, we can. All right. So, you can actually build AI agents for free using Quaza Alpha inside NA10. To use NA10 for free, all you need to do is host this locally. And then to use open router with qua alpha then all you need to do is just go to your AI agent over here plug that information right here and then you can use the credentials directly. All right so you can see open route alpha and then if we hit save on that open up the chat and say are you working just going to make sure the AI agent has the system message right here. Try that again. See if that works. If not, what you can also do if that's just playing up or something like that, then you can also switch to another different model here. So, let's switch to Quen 32B instead. And there we go. Boom. All right. So, when a chat message is received using the AI agent and then inside open router, select your credentials and add Qua Alpha and you are good to go. As you can see right here, it worked perfectly. I think the first time it didn't work cuz I didn't set up that was one of my team that managed the AI agent. But if you just bear in mind if your team sets up other AI agents inside 10 doesn't seem to work when you plug in your open router details. You can see that works perfectly now right here with the AI agent created in a couple of minutes. And then if you want to edit what it does etc you can just insert the system message right there. But basically between using root code NA10 and also open router chat you can basically build whatever you want using quaz alpha. It must be a new release from chat GBT I would say or open AI but it's going to be interesting to see when it gets announced and what's coming next. All right so thanks so much for

watching. If you want to get access to my AO profit boardroom with all my best agents automations etc workflows feel free to get that link in the comments description. price is going up this month, so make sure you sign up now before you miss out. And it comes with all my best SPS, tips, templates, workflows, etc. And this gets updated pretty much daily as you can see. On top of that, also comes with weekly Q& As's focus on making you more money and saving time with AI. So you can jump on the weekly live calls here, ask any questions you have. And inside the community, we have 694 members, right? So you can post in here, ask people for help, and then anytime you get stuck with AI, feel free to post directly here, right? So, thanks so much for watching. Feel free to get that inside the link in the comments description. And additionally, if you want to get a free one-to-one SEO strategy session that shows you how retail websites from zero to 145,000 vis generate hundreds of thousands of dollars in sales and more toilet, feel free to get that. Link in the comments description. Appreciate you watching. Bye-bye.