Transcription
Hey there everyone. So finally Kim has released their thinking model, the Kimmy K2 thinking. And do we finally have a competitor to Sonnet 4.5? Let's directly get into the coding competition. I'll be talking about Kimmy, its reasoning capabilities in another video, and what is special about it. But in this video, we'll be directly heading into the coding competition.
So, if you now talk about the pricing, you can see Moonshot Kim K2 on Open Router mentions that it is 60 cents per million input tokens and $2.50 per million output tokens. On the other hand, our competitor Sonnet 4.5 is $3 per million input tokens and $15 per million output tokens. So, it is approximately six times pricier than Kim K2 thinking. But still, will Kim K2 be actually able to defeat Sonnet 4.5? Let's find it out.
So, the first question is: Can you draw a dragon in a web page using HTML, CSS, and JS? Show a side view of the dragon and give your best implementation in one single code. So, you guys know it. It is the dragon question. Let's see how they do.
Okay everyone, we have got the responses from both models. Towards my right, I have Sonnet 4.5, and towards my left, I have Kim K2 thinking. If you talk about Sonnet 4.5, you can see it is clearly distinguishable as a dragon. We have the horns, eyes, nose, head, neck, body, wings, and everything. It is very close to a perfect response. Talking about Kim K2, uh, it is not that perfect because, uh, you can especially look at these horns. They are just, uh, in the air or on the wings. The face is implemented correctly. You can have a look at the eyes and stuff, and they are kind of following me. So, that is there. But comparatively, I think Sonnet 4.5 is better. So, this is a win for Sonnet 4.5. Let's move on to the next problem.
So, the next question is: Build the absolute best implementation of Super Mario Bros. 1985, the original and the most iconic game for Nintendo Entertainment System. It has to be built in a web page without using any external assets and in one single code. Let's send this and let's see what they come up with.
Okay everyone, we have got the responses. Let's start off with Sonnet 4.5. So, this is the game it has come up with. Let's hit space to start. And this is what we see. I think it has audio as well. So, let me enable that. Okay. So, you can see we have different sounds. We have powerups. We are able to kill the enemies. And we have the flag as well. If you cross it, you can see it says "Level Complete. Thanks for playing." Absolutely awesome response from Sonnet 4.5, as always. Let's keep this aside and let's come on to Kimmy. So, we'll preview this in a full screen. And this is what we have right off the back. You can see it looks a bit basic, and, uh, I don't see any kind of enemies or any obstacles or ways to have coins, anything like that. We have an implementation for cloud, but other than that, there's nothing else happening. Let's see. We'll continue to see if there is anything different. Yeah, I think we are close to the end, and there is nothing else. Also, there is no sound. So, it is a terrible response, I have to say. So, it is a clear win again for Sonnet 4.5. I mean, this model is very hard to beat. Let's move on to the next question.
Now we are asking them to create a stunning and detailed page about the solar system. It has to use scrolly telling. This is very similar to what you see in websites like Apple. You know, when you scroll, you see different animations happening, phones and devices moving here and there, and stuff like that. So, they have to implement such kind of movements, but this time with planets and objects in the solar system. This will give us a good idea of the web design capabilities of these models. So, let's send this to both of them and let's see what they come up with.
Okay, we have got the responses. Let's start off with Sonnet 4.5, as always. So, it says "Journey through the solar system." Let's scroll this, and we have animation for the sun. We have different points. Uh, I have seen that, you know, Anthropic models usually go with SVG implementations like this instead of using an image. You can see the sun is kind of floating. That animation is nice. Now we have Mercury, Venus, Earth, Mars, Jupiter, Saturn, Uranus, Neptune. Also, when I move the cursor, it is kind of creating a star-like trail. All in all, it's a good response, but we'll have to look at Kimmy's response after that. We can make a decision. So, let's preview Kimmy's response. It says, "Journey through the solar system. Scroll to explore our cosmic neighborhood." So, this thing is not implemented as good as Sonnet did, but let's scroll. Okay, it says the sun. And look at this. So, if I scroll, you know, one and each point is revealed one by one, which I like. Next, Mercury, you can see the planet size kind of increases, and each point reveals itself. Then Venus, Earth, Mars, Jupiter, Saturn. Saturn image, it was not able to fetch. Uranus, Neptune. So, the introductory part was not that good, but the rest of the things are actually nice, and it was able to fetch the images. And if I talk about visual appearance, I would say that actually, I think, uh, you know, instead of using SVG images like that, it actually used the correct images, and also the animations and background are very nice instead of just being plain black over here. So, I would say that Kim K2 is slightly better over here. But you may differ. It's up to you. Let's move on to the next question.
Now we are asking them to create the best implementation of the game Microsoft Flight Simulator. You can have a look at the prompt. Let's send this and let's see what they come up with.
Okay, we have the responses from both models. Let's start off with Sonnet 4.5, and it loads, and this is what we see. We are clearly able to identify this as a flight, and if I hit shift, you can see the blades kind of rotate. We can also change the camera view, and it looks absolutely insane. I love it. You can have a look at this free mode as well. The environment is very nice, but the problem is that, you know, I'm not able to move it forward. I've tried it. You can see I'm able to rotate here and there, but it is not moving forward. So, this needs to be fixed. Having said that, let's keep this aside and let's try Kimmy K2 thinking. So, we'll preview this, and it is loading. Okay. So, we do see something, and I'm able to move it. It is a bit more aggressive than what it was with Sonnet 4.5, but, uh, it is kind of falling downwards, I guess, and I'm again not able to move forward. If you talk about the plane, it is not as detailed and as coherent as Sonnet 4.5. So, I think we'll have to come up with a second iteration for both of them because, you know, the flights are not moving forward. So, let's come up with the second iteration.
Okay everyone, we have got the updated response. So, let's preview Sonnet 4.5 first. So, it's loading, and you can see we are slightly able to move here and there, but, uh, after a while, it just vanishes, and, you know, I'm trying to move left and right, but it's not able to move. It's not able to move freely. So, we have this issue over here. Let's keep this aside and let's try Kim K2 now. We'll preview this, and okay, I'm I'm able to move here and there, which is nice. I mean, I'm I'm able to rotate, guys, but, uh, you can see the flight is kind of moving in one direction. It's kind of going down. So, maybe I was wrong. It is just able to rotate, but it has the same issue that Claude has. So, again, Kimmy K2 was also not able to do it. So, I'll give a fail to both of them. So, let's see.
Now they have to actually build a trading dashboard with all the tools and tick-by-tick updating chart. So, this is also a very popular use case. Let's see how they do.
Okay everyone, we have got the responses. Let's start off with Sonnet. So, we'll preview this in a whole screen, and this is what we see. So, it is a proper chart, but there is one issue. I guess this, you can see we have two bars instead of one. That shouldn't have happened. Let's see if different chart types work. Line chart, area chart, OLC works. Let's see if different time frames work. Okay, different time frames is also working. Let's see if indicators work. Okay, indicators is also working. So, it is an okay response. The only issue being, you know, having this two candles instead of one. So, that is an issue. Let's keep this aside and let's try Kimmy now. So, we'll preview this, and this is what we have. Uh, it is kind of a distributed and not the way a real candlestick chart would look. Let's see if different chart types work. So, line chart, area chart, Hikinashi. Okay, I'm clicking different chart types, but nothing is happening. Okay, we are able to change the time frames. That is nice. You can see if I change the time frames, it is changing. We have the ability to draw, and you can see I'm able to draw lines over here. But I'm not sure how to exit. Okay, now I'm able to exit. We can place orders. If I hit buy, it needs quantity. So, we'll just enter 100 and hit buy. It says "Order placed." And we are able to see the order down there. So, it is nice, but the actual look of the candles on also, you know, this tick-by-tick update that is happening. You don't see the candles getting updated. Other than that, the rest of the things look kind of good, but, uh, I think Sonnet is slightly better because the chart looks more real, has more things functioning, especially the chart types. So, I'll give this win to Sonnet 4.5. Let's move on to the next one.
Now we are asking them to create the best implementation of the game Angry Birds. Let's see what they do.
Okay everyone, we have got the response. So, let's look at Sonnet's response first. So, we'll preview this. Okay, we do have three birds. You can have a look at the angry birds. And if I pull back, we are able to do it. We also have a trail where it will go. Let's release it. And okay, the slingshot power is actually less. I have seen this many times. Usually, the slingshot power is actually less. I don't know why it's the case. Anyways, this is the last board. Let's try to hit it. Okay, so we're not able to finish it, but it needs improvement. Let's keep it aside and let's try Kimmy now. Let's preview this. And this is what we have. I don't see any kind of birds, any platforms, or anything like that. So, I don't know what to say about it. So, I think it's a clear fail. And this win is for Sonnet 4.5.
Okay. So, this is something new that I am going to try. The prompt is: Generate the most epic music with musical notations alongside it that you can in HTML, CSS, and JS and in one single code. Let's see. I'm kind of excited.
Okay, we have the responses. Let's start off with Sonnet first. So, I'll play the whole audio. Be patient and have a listen. Let's play this. [Applause] [Music] [Applause] [Music] That was nice. I actually loved it. It had a soothing kind of feel. Let's listen what Kim has come up with. So, we'll preview this and we'll hit play. Okay, nothing is audible now. Now, it's audible. Heat. [Music] Heat. [Music] Heat. [Music] Heat. [Music] [Music] Okay, it was kind of different. I will not say it is very bad, but I think, uh, Sonnet actually did a very good job. I think, uh, Kimmy had some elements which were less audible. I was able to hear it through my headphones, but, uh, Sonnet's was more audible and also it was more had more feel to it. That is my perspective. Music is subjective, as you guys know. But I'll give this win to Sonnet 4.5. Let's move on to the next one.
Okay. So, now we are doing some Python-based problems. The first question is: Write a Python code for a hollow circle with an opening. There are two balls inside it, and whenever a ball exits, two new balls appear. So, this is the animation that they have to build. Let's see what they do.
Okay everyone, we have the responses from both models. Let's start off with Sonnet. Now we'll paste the code and let's run it. Okay, we do see two balls. You can also hear the audio. Okay, I think one ball exits now, and we have two new balls. Okay, as we have new balls, you can see we have an issue. The balls are kind of getting clumsy down at the bottom. So, this shouldn't have happened. So, let's look at Kimmy's response. If Kimmy's response doesn't have that clumsy kind of issue, then I'll give it a win. But it has to work in the first place first. So, okay, quantum escape. Let's hit space. Okay, we do have two balls. The looks are not that good, I must say. Let's see. One ball is exiting. Okay, we have two new balls. So, now you can see over here as well, we have this clumsy kind of issue. You can see all those balls are down there. So, both are not perfect, but, uh, one is not better over the other. So, I'll give the point to both of them. Let's move on to the next one.
Next up, write a Python program that simulates Earth's rotation. So, they have to show the Earth rotating, and if possible, they should also have a moon revolving. Let's see what they come up with.
Okay everyone, we have got the responses. Let's start off with Sonnet 4.5. We'll paste the code and we'll run it. And we do see a black screen, but it didn't work. Let me try it again. I'll run this, and you can see this is a black screen, and it just crashed. So, we have an error over here. We'll copy this. We'll paste it, and meanwhile, let's see what Kimmy has come up with. We'll paste the code and we'll run it. And again, we have a black screen, and it closed. So, Kimmy was also not able to do it. So, we'll share their errors and try to fix it. We'll come up with the second iteration.
Okay, we have got the updated code. So, let's try this. We'll paste the code and we'll run it. Okay, we finally see something. I think this is representative of Earth, and we have the moon over there. It is not one of the best implementations that Sonnet has come up with, but, uh, okay, let's see what Kimmy has come up with. Let's copy the code. We'll paste it and we'll run it. Okay. So, this is what it has come up with. There is no moon though. I'm not able to clearly distinguish this as Earth. Okay. Okay. If I zoom out, I can see the moon. So, again, I think both have given a very close response. I think I'll give a win to both of them.
So, next up, we are asking to generate the game Pac-Man, released by Namco in 1980. One of the most successful video games of all time. Let's see what they do.
Okay everyone, we have the response. So, let's start off with Sonnet 4.5. We'll paste the code and we'll run it. Okay, it says "Pac-Man. Press any key to start." And we're able to play, yet able to take the pellets. But I think it is very sped up. We are able to eat the ghosts as well, but it is very sped up. And I've seen all the functionalities are working, which is nice. But yeah, it is very sped up. That can be fixed. Let's look at Kimmy's response. So, we'll copy the code. We'll paste it and we'll run it. Okay, it says "Pac-Man. Space to start." So, it directly says "Game Over." I don't know. I'm not able to play it. So, I think we know who is the winner over here. I think Sonnet 4.5 is better. So, we'll give this win to Sonnet 4.5.
With that being said, we are done with our testing. What can we say it? So, it's very clear that when it comes to front end, there is no competition. Sonnet clearly beats K2. We did a total of seven questions, and out of them, in five of them, Sonnet 4.5 was better, and only in one of them, Kim K2 was better. Talking about Python-based problems, we did three questions, and it was kind of a neck and neck between both of them. But out of three questions, Sonnet 4.5 gets three points, whereas Kim K2 gets two points. So, the difference is very big in front-end design, but when it comes to back-end, it is not very big, but Sonnet 4.5 still remains the king. I think, uh, Gemini 3 will be the only one which will be actually able to defeat Sonnet 4.5. Having said that, I hope this information is useful to you and you got some value out of it. If you like this video, please do hype this video so that more and more people are able to watch this. With that being said, I'll catch you guys in the next.