Welcome everyone. Imagine being able to use GPT5, Claude 4.5, Gemini 2.5 Pro, and Gro 4. Plus, generate stunning images and videos with Nano Banana and VO3, and even build websites also like Lovable. All completely free and unlimited. Sounds crazy, right? But it's real. It's all possible on lmarina. 100% legal public AI research site where the world's top models compete and you get free direct access to test them yourself. This is the lmrena.ai site. You can see it says find the best AI for you. That basically means you can try out any model you want totally free and unlimited. Now I'll open the leaderboard page. Here you can see which AI models are performing the best ranked based on real user votes from around the world. For text generation, Gemini 2.5 Pro has the highest score in vision models. Once again, Gemini 2.5 Pro is number one for text to image. Hunuon image 3.0 is leading in the web search category. Gro 4 takes the top spot. And for both text to video and image to video, V3 is winning the votes. Once you start using LM Arena AI, you can also join in and start your own votes. If you want to save your chat history, you'll need to log in to the site. LMA Marina AI also works without login, but in that case, your chat history won't be saved. Click the battle menu. The first option says battle with two anonymous models. When you select this, you can start a battle between two AI models and compare their answers. There's one more option called direct chat. Use this to pick the exact model you want. This is where you can use paid models for free right in the browser. By default, Gemini 2.5 Pro is selected, but you can switch to others easily. Available models include Cloud 4.1 and Cloud 4.5 Pro, GPT5, 03, Q3 Max, and GLM 4.6. You'll also see advanced models like Deepseek Advance, and more experimental options. Just open the model drop-down, pick one, type your prompt in chat. That's it. Now, I'm going to ask something. This is my prompt. I'm asking for product ideas for a solo developer. I'm using the Gemini Pro model here. So, the response will be generated by Gemini Pro itself. Boom. After just a few seconds, the Gemini Pro model gives the top five ideas, all generated instantly. Everything here is coming straight from the Pro model and it's awesome. Now, let's check another model. I'll choose the Clawed Sonnet advanced model. I'll use the same prompt again asking for product ideas for a solo developer. And here we go. Claude 4.5 also gives amazing results. The ideas are short, clean, and really practical. We can also use web search models. You can see options like Gro 4 fast and perplexity sonar pro. There are also search variants for 03, GPD5, and claude. I'll pick Perplexity Sonar Pro for this test. I'm going to ask it about creating a blog with Nex.js16. Nex.js16 is a recent release, so this will check if the web search model returns up-to-date info. After a few seconds, it returns an answer. Everything here is correct. It's specifically about Nex.js16 features. So, if you need the latest realtime content, use the web search models for up-to-date results. Now let's move to the image generation models. Click the image option and open the drop-down. You'll see the Google Nano Banana model listed there along with other advanced models. I'll use nano banana for this test. I'm going to upload a photo of a person and ask it to change the dress color. For image generation, you must log in to the site first. After logging in, I'll enter the same prompt again and hit submit and wait for a few seconds. Boom. See that Nano Banana worked perfectly. It changed only the dress color without affecting the face or body. Everything else stayed the same. That's really impressive. Now, let's check the Hyuan model. This one is the top ranked model for image generation. I'll ask it to create a random anime character. And there it is. It just created a beautiful anime girl character. Looks awesome. One more cool thing. We can even start a battle between models. Here, I'll type the same prompt. And now two AI assistants will respond at the same time. You can compare both response and pick the one you like the best. That's awesome, right? So, the two assistants have finished creating the anime characters. One model made a boy character and the other created a girl character. Both look great, but now it's your turn to decide. You can give your vote and choose which one you think is better. It's not just for images. You can also ask real coding related questions. I asked both models for Nex.js JS15 app router code and both AIS gave really nice answers. You can even manually combine their outputs to create an even better result and start to work on next. Using LM Arena, you can even build full websites just like popular tools such as Lovable or Bolt. How cool is that? Okay, check this out. After logging out, I found a hidden option. Click this little icon and it opens a brand new subdomain, web.lmarina.ai. If you don't see the icon, no problem. You can manually open it, too. Just add web before the main URL. Here, you can type your own prompt or pick one of the suggested prompts. I'll choose the Twitter clone UI prompt. Now, two AI agents start working on it. They each generate a version of the site. When they finish, you pick one to use as the final result. For me, the first agent worked perfectly. It generated a Twitter profile style UI. The second agent had a few issues. If you want to fix those issues, just continue the chat and ask the agent to correct the errors. Awesome. Finally, let me show you how to use the V3 video generation model. For this, we need to join the LM Arena Discord channel. Once you have joined, look for the channel called generate a video. Just click it. Here you'll see that other members have already created lots of videos. If you want to make your own, simply enter your prompt and hit submit. The model will start generating your video. One small thing to note, each member can send only one message every 5 minutes. So be patient while your video is processing. So that's LM Arena AI, one platform where you can try the world's most powerful AI models, all completely free and unlimited. It's fast, fun, and super useful for creators, developers, and anyone who loves AI. If you found this helpful, make sure to like, subscribe, and stay tuned. I've got more amazing AI tools coming soon. Thanks for watching, and see you in the next one.
Get free YouTube transcripts with timestamps, translation, and download options.
Transcript content is sourced from YouTube's auto-generated captions or AI transcription. All video content belongs to the original creators. Terms of Service · DMCA Contact