[Music] [Applause] hi welcome to another video if you guys remember I showed you WebDev Arena when it first came out it allowed you to use a bunch of models for literally free and without any rate limits which was insanely cool at the very least recently I saw that their interface has been revamped and the leaderboard is also now available with a ton of new models as well which is kind of amazing now if you don't know about WebDev Arena it is pretty much like clawed artifacts or OpenAI canvas or anything like that which you've seen but because this is by LMSYS it also has a leaderboard where you can see which models are the best basically you can send in a prompt and then you'll get two responses of code from two disguised models then you can view the code and you can also preview it to see which has a better generation then you can vote for which is better like the left one the right one or mark them as a tie based on that the model score will improve allowing it to climb the leaderboards this is fully free to use and you don't even need an account which is great so let me show you how you can use it and how it actually works but before we do that let me tell you about Ninja Tools ninja Tools is an AI platform that combines all the best AI models and experiences at one place it allows you to save over $600 per year compared to having separate subscriptions you get access to Claude 3. 7 Sonnet GPT40 Gemini and a ton of others models in one subscription you even get some more cool options like AI video generation image generation music generation and document chats you can also use their playground to compare multiple AI responses at once the best part is that it just starts from $11 per month that gives you more than 1,000 chat messages 30 AI image generation and five music generation while there is also some even more advanced plans if you need them also make sure to use my coupon code AI code king 20 to get an additional 20% off make sure to check ninja tools out and save some money on your subscription while you're at it now back to the video first of all you can just come to this new webdev arena site and here you'll see this interface on the left you have the sidebar where you can see the new chat option and the battle option both do the same thing open this interface the leaderboard will show you which models are the best at coding like Claude 3. 7 Sonnet is currently at the top followed by 3.
5 sonnet and then we have Deepseek which is all kind of great it even has Gro 3 which is kind of insane to see anyway then you have the main things here here you can see that this is the prompt box where you can send the prompt of what you want to generate or anything like that which is kind of amazing you can also use the surprise me option to generate something random or you also have some prompts here so let's write a simple prompt to create a simple image cropper tool app once you've written it you can just send in the prompt and then you'll see that both the models over here start generating the code you can see them being streamed here generally the responses are extremely fast so after just waiting a bit you'll see that the code is now generated and it looks good you can also see the preview in the block thing it uses E2B for the preview so you can also take the URL and share it publicly which is great anyway you can see the generations from both models here both look good but this one is better over here so now you can vote for which you like best here or you can do another thing and that's the option to send in another prompt and edit it as needed for example we ask it to make the colors red so let's type in the prompt once you've typed it in you can just send in the prompt and it will again start rectifying the code to accommodate the thing you asked for this will again take a bit because things need to be streamed again if we wait a bit you'll see that the code is now generated and you can see the previews update regarding that once you like either of them you can just vote for one or you can also tie them if you think both are good or both are bad once you do that you'll see the exact model you voted for and which one you didn't this is pretty great for sure even after voting you can keep iterating on the stuff as well so like we can ask it to do something again let's ask it to add a title saying king once you've written the prompt you can just send it and then you'll see that it again starts editing the code and everything which is pretty great once it's done you'll see that it does what we asked it to do which is great if we wait a bit you'll see that it's now done and you can see that it did what we asked it to do this is also super cool as it's a proper usable tool that you can use to generate front end and other stuff while helping the leaderboards as well you can also download the code and copy it if needed so that's super cool as well i think this is great to use because it doesn't cost even a scent and you don't need to create an account which is also great another thing is that it only generates React code it doesn't generate anything else which is a little bad for sure but still it's great for generating some basic stuff like you do with clawed artifacts or canvas so that's super cool also you can view the other text that the LLM provides in it by closing the preview i think these things are really a good way to test these LLMs in practical use cases because not everyone just wants to generate simple text most of the time we use it for coding and stuff and I think that this leaderboard will be great for sure as it will allow us to see which model is really good at coding real things versus which are not as good also if you're wondering which models it has it has a bunch of them mostly you'll see the new experimental models like Polus which is rumored to be Llama 4 or Claude 3. 7 Sonnet Quen 2.
