Transcription
[Music] [Applause]
Hi, welcome to another video. So, a new stealth model has launched on multiple platforms known as code supernova. This is a new stealth model, possibly by XAI because I have some concrete evidence pointing it to be a Grock model.
It is a multimodal model as well, but it only supports image inputs, which is what most of you would use. But I also sometimes use video inputs, which are not yet supported. Still, that's fine. It has a 200,000 context window combined. So about 128,000 input and 64,000 output possibly. So there's that.
It's also a reasoning model allowing you to set it between low, medium, and high. It's unlimited without any cap on Kylo code, and there's no throttling either. It's currently super fast and pretty good as well. They say that it's great at visual context understanding and can interpret screenshots, diagrams, and mockups alongside text. This helps the model understand requirements when building. Especially handy for front-end heavy projects. You can share screenshots of error messages, console outputs, or unexpected UI behavior to get more accurate debugging help. It's also good for design to code workflows, allowing you to add mockups, wireframes, or even handdrawn sketches and translate them into code. Basically the focus seems to be mostly on multimodality here.
I have tested it on the four agentic questions and not the kingbench questions because it is only available on coding agents and not the API making it not possible to test it fairly as the agents have a ton of system prompting which would not be fair for other models. So let's look at it.
But before proceeding, let me tell you about Ninja Chat. Ninja is an all-in-one AI platform where for just $11 per month, you get access to top AI models like GPT40, Claude 4, Sonnet, and Gemini 2.5 Pro, all in one place. I've been using Gemini for quick research. But what's really cool is their AI playground where you can compare responses from different models side by side. Their mind map generator is a gamecher for organizing complex ideas as well. The basic plan gives you 1,000 messages, 30 images, and five videos monthly with higher tiers available. If you need more, use my code king 25 for 25% off any plan or king 40 yearly for 40% off annual subscriptions. Check the link in description to try it yourself.
Now, back to the video. I tested it with kilo code. You can just install it easily if you haven't done that. And then just go here and select the supernova model and you should be good to go. I was looking at the request logs to find out what the model is. And I'm really sure that this is a Grock model because it outputs this numbum sources used which is something exclusive to the Grock models API. None of the other providers or models have this element. So I'm really sure that this is a Grock model. probably a bigger version of Grock code based on my testing.
So, let's look at the testing results. The first question that I tried it with is to ask it to make me a movie tracker app and it is good. It looks really good, but the container for the movies overflows from the div, which is not great. It also loads each and every movie without any limit which is not needed at all. It overflows the memory as well as spams the API which makes it not a very usable generation. Though the inner pages and everything look super good but all of them are not in the perfect territory. Like it still leaves the top bar in inner pages which should have been removed. Comparing it with what you get in Claude Code, I think that Claude Code Gen is better because the UI, although not as visually good, is cohesive and usable in the first try, which to be honest is great. GPT5 codecs and stuff are also better, but this Supernova model is not bad either. It's really good and it's super fast as well.
Moving to the next question of making a goateui calculator, then it ain't good. It can't really align the elements well. So, this is a fail. But the go game where I asked it to edit an FPS game and add a step counter and a life bar that is affected when a player jumps. This generation was a great one. It is awesome at Godo. The previous Grock codef fast was also fine at godo. However, this is much better and you can see that although the UI could be slightly better, it's perfectly usable and it worked in one shot which is quite awesome. However, it still struggles a lot on longer tasks.
In the open code test where I asked it to edit the open code repo in order to add an SVG generation modal, it failed badly and actually started to glitch. You can see here that it just started to say let me check a bunch of times and ultimately Kylo code itself detected the glitching and stopped it. This is something that I myself have encountered with grock code fast as well. In many long generations, it also failed and glitched and this happens here too. So yeah, not as great. I would still say that GLM4.5 is better and the closest to Sonnet and if the glitching and model performance get a little better, then I can surely recommend it. It is a good model, but it's just not there yet. It needs to be a bit more reliable. The speed here is great, though. Still, I would recommend GLM4.5 for everyday usage, as that is better, especially in agentic contraptions. It does come with multimodality though, which is really something that I myself need a lot of the time, and it's pretty good at that as well. I think this might be a better version of Grock code with multimodality included. So yeah, there's that.
You can also use this model in things like rue clin and open code and the performance is very similar. It's nothing extraordinary yet. I'm pretty sure this is a better version of gro code with multimodality or something similar. That's pretty much it. go ahead and use it and let me know what you guys think about this as well. I like it and let's see what it turns out to be. Considering that this is a firstg coding focused model from a relatively new lab like Grock, it is a really good feat for sure. It is super fast as well. Overall, it's pretty cool.
Anyway, share your thoughts below and subscribe to the channel. You can also donate via super thanks option or join the channel as well and get some perks. I'll see you in the next video. Bye.
[Music]