NEW Google Gemma 3 AI Update (FREE!) 🤯

Gemma is basically a new API that just came out from Google. So, pretty much just came out today and this is announced as the most capable model you can run on a single GPU or TPU, right? And a few different versions of Gemma, but this is Gemma 3, which is more like a lightweight version of the models built that power Gemini 2. 0. Now, you can actually run this locally. I'll show you how to do that in a second. It's also available in like multiple languages. So you can create AI with

like advanced text and visual reasoning capabilities and you can also test out online but you can see how it's performing here on the chatbot arena elo score. So I'll show you the leaderboards in terms of the LM arena in a second but essentially you can see here that it's actually just behind Deep Sea R1 but it's outperforming Deep Seek version 3. And what's crazy is it's actually outperforming chat 03 mini which is wild when you think about it. So this is a free model and it's super lightweight whereas you got 03 mini being outperformed by it deep sea version 3 llama 3 m large and gemma 2 all being outperformed by quite a way by Gemma 3. Now I always take these benchmarks with a little bit of a pinch of salt but you get the point and we'll run through some of these examples. So, let me pull this up for you right here. Yeah, that's the other thing as well. It's like multimodal, right? So, it actually has vision, which is pretty useful. So, you can like understand, for example, like this is a picture of looks like an aircon remote and obviously it's in I think it's in Japanese or Chinese. And so, you can take a picture, upload it to Gemma 3, and then based on that image, it can just translate it for you. So, it's actually in Japanese. Yeah. And

then if you want to get access to this locally, you can get that set up with Llama, right? So if we go to O Lama over here, you want to make sure you download O Lama. Then make sure you have this running. So we're just going to open up O Lama. So you can see it's running in the top right here. So we've got the icon brick. And then if we go over to the model section, the most recently uploaded model is Gemma 3. So you got a few different versions of it. you've got, for example, on 4B, 12B, 27B. Okay, obviously 27B would be the most powerful, but it's also a really big model. It's going to require a lot of power for you if you're running it offline. And I'll start with the really basic one, which is the 1B method. You can also use this online. I'll show you that in a second. But if we copy this

then we're going to go over to terminal. And then from here, we're just going to paste in OAMA space run space GEMA 3 1B, right? And so what that's going to do is start installing it for us like you can see and we can just wait for that to load. In the minute we got a

couple of questions here. So Sun Sundep was asking, I got hit by the core update in December, how to recover. Feel free to book an SEO strategy session about that or join the SEO elite circle and we can help you. People asking how to get madness. Just check out my video on open manus or watch the live stream. So why is that not working? This model is not supported by OA. Maybe I need to upgrade OAM. So I'm just going to do that right now. Let's log in here and we'll download the latest version. But that's useful to know as well for the sake of this video so that you can understand if you've already got a pre-installed version of Lama, you just need to get OAMA 3, it looks like. So, we're going to open up that. Make sure you got it downloaded. Move it to applications. Then, it's going to start installing it. And then we just have to run through the install section. And then we'll copy this. And hopefully that works. So, if we go back to Gemini now instead inside the models, we'll grab the terminal command for 1B. Grab that. Go back to terminal. Hope that works. Look at that. Should be working. All right, let's test this out. So, say what model are you? And it says I'm Gemma a large language model. That is really fast as well. It's so lightweight. All right. So, for example, say code HTML cost calculator for SEO. Look how fast that is for AI running offline. Bear in mind, this is like some of the best technology in the world. It just runs locally for you like that. So quick. That's absolutely awesome. Here we go. All right, let's check this actually works. So, grab the content like so. And then we shall copy that. Let's preview it. Preview this bad boy. All right, we'll go over to liveweave. com. This is like a really good website just for previewing. and HTML. And we have the HTML ready to preview like so. Boom. Wow, that was easy, wasn't it? I don't know if it's actually going to work. We can test it, but pretty easy and simple to set up. I don't think it's not the best design in the world, honestly. But that's using the 1B model, which is pretty basic. So, that's how to install Gemma locally using 1B. You can go all the way up to 27B. Honestly, for most people, that's I assume that's not going to work very well, but haven't tested it. So, maybe you can test out

yourself. And then if we go to LM arena and the leaderboard here, we can see the new leaderboards for the LM arena. And Gemma is performing pretty well here. So you can see here Gemini is just right up there with Quen 2. 5 Max. It's ranking joint ninth along with 01 preview and 03 mini high. Now bear in mind like 01 preview and 03 mini high really powerful reasoning models. And also you can't run these locally, right? So you can't run, for example, 01 preview on a Lama, whereas for example, Gemma 3 like you've seen today. You can install it, run it locally. Pretty powerful. That's the 27b model, but still pretty cool. And then if you want to test this out, you can go to aistudio. google. com and you can switch between the different models. So you can see here, Gemma 3. You've got the older version, so like Gemma 2 as well, but we'll stick with Gemma 327B. And then test this out. So I'm

going to say okay code and SEO cost calculator in HTML for me make it beautifully designed if I can. And then we'll hit enter. And again this is free. All this stuff is free which is pretty awesome. And you can also get a free API as well. So if you want to get a free API key just go to get API key inside a studio. google. com. Pretty simple and easy to do. Wait for that to load. And if you want to know okay what are the key benchmarks what do you need to know about this and stuff like that. So I actually got perplexity to do a deep research on this and essentially it has number one has 128k context window. It's pretty good for offline model for a localized version. You can see the differences here. So, for example, if you install Gemini 3 1B, it can only read words. 4B, read words and see pictures. 12b, more pictures. 27B, do everything pretty well. All right, we like that. And then let's go back and see what we got here. So, let's copy this. Plug it in. We have the cost calculator. That actually looks a lot better than the 1B version. And it works perfectly. like you can see so you can create tools with it. Pretty simple and easy. What I actually do, I want to test this out is I'm going to take this prompt that we used inside Manas and just see if it works inside Gemma 3. So let's plug this prompt in which is for Julian Gold agency create a landing page. Here's some testimonials to insert etc. code it out and see how it compares versus the custom GPT page. Now, if you want to see an example of what the Manis output was like, let's pull this up. So, we got the subdomain right here and then we can compare this page which is built with Manis versus the outputs from Gemma 3 and just see how it performs. So, it's writing the content here. I actually quite like the content it's writing. Stop doing work. Let AI do it for you. Look, you're busy. You got a business to run. So, I like the fact that number one, it can code it, but number two, it can actually write like pretty good copy. Honestly, that's going to be better than most freelancers that you hire, unless they're using like the prompts that I use for sales copyrightiting. And then it's building out like you can see. One thing you got to be really careful of here is like you see how inside here is like Julian and his team are absolute wizards. They built a custom GBT that's already saving me hours per week. So you got to be careful because like quite often when you're building out sales pages with Gemma 3 or any sort of other AI, it will hallucinate testimonials and it's a nightmare for that. So you got to be really careful of that. And I should have included that inside the prompt for this particular page, but you can always go back and forth a bit later. We'll grab the preview in a second as well. By the way, if you haven't already tried this, you can also stream real time inside AI Studio. And that's pretty cool because inside AI Studio, that means, let me terminate this. That means that you can, for example, share your screen, talk to the AI, can see everything. You can use it like a personal assistant. So, let's grab the HTML from here. We'll plug that in. And that is not bad at all. Look at that. It's not probably not as good as Manis, but it's pretty decent. It's embedded the testimonials as well, which is nice. Obviously, you would get rid of the fake testimonials that it's written for some reason. And then, if we click on that, does it take us to the right link? Absolutely. So, that's pretty cool. Let's see if we can actually take

a screenshot of it now with Gemma 3. So, we'll grab the screenshot, copy that image, and then from here, we'll go over to Gemma 3. We'll paste in the image. Oh, it doesn't support images. Look at that. That's interesting. I thought it was supposed to be multimodal. Maybe you have to upload a file. Nah, look at that. Doesn't support images. I thought it did. Thought it was supposed to. Maybe it just doesn't on this version of things. So, that's good to know as well that you actually can't. But yeah, I like the design of the website and that sort of thing. Let's just go back here. I'm sure it said multimodal. Yeah. So, it was saying like you can it's got multimodality here. Integrated vision encoder, but it doesn't seem to work directly inside AI Studio. Maybe it's just inside the chat directly. Interesting. I wonder if we go to Is that going to be available in Gemini as well? Should be on the paid plan for Gemini. Can't see it in there. So, it's not built into Gemini, but it is built into AI Studio. And then also, of course, you can get it on Lama as well, I think. Also, if I'm not mistaken, you can try it inside the LM arena. So if we go to arena direct chat and then from here we'll type in Gemini. All these links are inside the AI profit bordering by the way if you want this stuff. So we'll switch to Gemma 327B as you can see. And then can we insert our image here? Yes, we can my friends. All right. So that's how you insert images inside Gemini. So just go to hugging face or host it locally and then from there you can start using it. And then I'm going to say, okay, redesign this page so it looks more beautiful. See the image attached. And then I'm going to grab the code from LiveWeave. Let's grab that code. Go back to LM Marina and I'll say improve this. Improve and rewrite all this HTML. So it looks more modern and sleek. Look at that text only model. What are we talking about here? I don't know whether it's it might maybe it's only the local model, but it doesn't seem to work inside LM Arena and AI Studio as well. So maybe I haven't tested it on the local version, but it seems like that multimodality it's promising isn't working from what I've seen so far. Interesting. Or maybe it's only the API that you can use that, but from what I've seen, mixed feelings about that. All right, let's see what it comes back with on the redesign in a second. Some people are asking like, is Gemma 3 as good as using for example like Manis? And I would say no. So, for example, if we compare side by side, this is the page from Gemma 3 like you can see. So, this is the landing page and we'll compare it side by side. So, let me pull up the window right here. We'll close the whiteboard down and then we'll open up Manis. All right. So, if we compare what we got back from like which landing page would you say is better designed? For sure, it's Manis. Right now, this is actually using Claude's API and I think that's why the content is much better. But if you had to compare Manis versus Gemma 3, this is not that bad, but it's not great. Whereas, if you look at the design from Gemma 3, it's more interesting. It's got more colors. It's got animations on the buttons for the CTA, which is where we want people to click and pay attention to. Is nicely designed for responsive. Really easy to read. Even the copyrightiting is nicer. So, for example, what's a custom GBT? I don't think that's as good as what can your custom GBT do for you with a nice little emoji right there. And then even like the icons, right? So if we have a look here in the here's what you get section, you can see how it's nicely formatted the icons here. It's got some nice copy. It's very benefit driven. Whereas this is super basic. This is just like some bullet points with a basic emoji and really basic copy there. Also, the content on this page right here is a lot more in detail and comprehensive, whereas the content inside Gemma 3's output. Not that interesting, not that great. We'll give it another stab at this. So, we've got the HTML back from Gemma 3 right here. So, we'll copy that and we'll see how it performs. Let's go back to LiveWave. Here we go. And honestly, not that much has changed, right? In fact, he's actually deleted the CTA button, which is where we wanted to send traffic. So, honestly, Gemma 3 nowhere near as good as Manis or Claude, but it's exciting that you can run it locally, which is the difference, right? You can't run Claude locally. You can't run OpenAI locally. You can run Open Manis, which is okay, but it's not as good as Manis AI. But yeah, and also Gemma 3, if you're using it inside AI Studio, it doesn't seem to be able to accept images. the same with LM Marina. So yeah, there you go. That's the full

breakdown. If you want all the links to it and all the resources, etc., it's inside the AI profit border. Just go to Saps, scroll down, and then go to Gemma 3 here. We also have some sections on a full Manus AI course. It's like a three-hour course in there. And also all my best AI agents, tons of workflows, AI automations, an awesome community of 447 members. You can ask for help at any time. And also you can jump on weekly Q& A calls with us. And if you can't make the Q& As's then you can join the Q& A call recordings like you can see right here and just watch them back. So all of the call recordings you can watch them back and learn from them as well. And we're just adding to this every single day. So every single day this is getting updated with the latest technology workflows and everything else. So feel free to get that link in the comments description.

description. And if you want to get a free one-to-one SEO strategy session that shows you how we take websites from zero to 145,000 visits a month and generate hundreds of thousands of dollars in sales on autopilot, feel free to get that on this free link building acceleration session. You'll get a free SEO domination plan. So you get a custom tailored link building plan to generate more lead sales and profits. You will discover his SEO link building. Answer any questions you have onetoone. Normally you're going to pay a consultant for SEO about $1,000 an hour if they're decent at what they do. We give this away for free. Link in the comments description. and you'll learn the best link building strategy for your website to get you more leads, customers, and sales to your site based on what's working for our clients and our happy customers like you can see right here. Feel free to get that link in the comments description. Appreciate you watching and I will see you on the next one. PS, if you're asking me for an invitation code for Manis, just type Julian Goldie open on YouTube and you can get all the details right there. Appreciate it as always. See you on the next one.