NEW Google Gemini 2.5 VS DeepSeek V3.1: Who Wins? 👀

Google Gemini 2. 5 versus Deepseek V3. 1. All right, so in the last 24 hours, Google Gemini have just released their latest model, which is Gemini 2. 5 Pro Experimental. As you can see right here, you can actually get free access to it. And you can also access Deepseek as well for free. So they have Deepseek v3. 1, and that is available to access at deepseat. com. You can also get free APIs for both of them and we'll be testing those out in a second. And basically what we're going to be doing today is just comparing them side by side. So these are the two latest AI updates, the two biggest ones in the game, I would say. And today we're going to be testing them out side by side and seeing which one performs the best. Now, the first

thing that we're going to kick off with because I like to give AI a challenge is by coding a game. All right. So, let me pull up an example right here. What I'm going to say to both of these models is make me a captivating endless runner game. Key instructions on the screen. P5JS scene, no HTML. I like pixelated dinosaurs and interesting crowds. Now, I could use it inside Google Gemini, but this is like the paid version, right? This is the advanced paid version. I want to make keep it fair. I want to keep it even Stevens on this comparison. So, what I'm going to do is I'm going to pull up Deep Seek V3 side by side versus Gemini 2. 5 Pro and compare which model performs the best. Now, one thing to note here is that Deepseek version 3. 1 is not a reasoning model, right? Whereas Gemini 2. 5 Pro is a reason model. So, there is a significant gap between them. But, I've been getting this request to compare them both side by side and compare the latest Reason models. So, let's see how they perform. going to hit

enter on both of them. So, we'll give them exactly the same prompt right here. The other thing that I was struggling with yesterday, I don't know if anyone else had the same problem as soon as this came out, was that actually with Deepseek V3. 1 after they released the new update, tons of people were on the website, right? And so, the servers were absolutely rinsed. It was super slow responding to us. So, let's see if it gives us better responses today and how it performs. Here's an example of the difference between a thinking model and a base model. So this base model with no thinking, it just jumps straight into answering the question. Whereas, for example, what you can see with Gemini 2. 5 Pro is that it really thinks things out before it does anything. Right? And it's going to spend about 10, 20, maybe even 30 seconds doing that before it finally comes up with the output. So we've got this back right here. We've got the example and it's finished the outputs before Deep Seek. So it's actually a lot faster. Bear in mind this is free, but Gemini 2. 5 Pro is a lot faster at getting back to us and typically smarter as well. And it gives us instructions on how to run it as well. So you can see how much slower Deep Seek V301 is. Now what we're going to do is we're going to test this out. So we're going to copy the JS from Google 2. 5 Pro Experimental. We'll plug that in and we'll open this up. All right. So we're going to hit preview cell. Open this bad boy up. See how it performs. And there we go. We've got the game ready to go. All right, that was pretty amazing. Incredibly easy to set up. Works perfectly. Actually quite a fun game as well. That's addictive. So, I'm going to close that now. And now we have the output back from Deepseek as well. So, let's copy that JavaScript over here. So, let's go into this. We'll plug it in. We'll open this up. So, this is the version from Deepseek instead. Let's see what we got. Not bad, actually. Which one would you prefer? I actually think that look at that's not bad. It's quite hard to see the score in the UI right there. But I do prefer the design. Like, and it's got more flying objects. For some reason, they put game over there when we didn't have a game over. It's got differences between night and dark, but it's broken. Look at that. I've not done anything and it just broke and said game over for no reason. It seems to switch between night and dark, like in between each night and day in between each go. But there's not many enemies coming here as well, so it's not as addictive as a game. All right, so I'm going to say Gemini 2. 5 Pro one there. But not bad at all. I'm impressed by Deep Seat V301's response right there. Pretty good stuff. Next up, what we're going to do

is we're going to compare this for the content creation. All right. So, I'm going to grab an example prompt from the AI profit boardroom and this is going to be my content prompt. All right. So, we're going to grab this. So, we'll copy this as you can see. So, this is one of my favorite prompts for writing SEO content. And we're going to run this through Gemini 2. 5 Pro, as you can see. And we'll put keyword equals SEO training in New York. And then we'll do the same inside DeepC V3. And we'll see which one creates the best content. One of the strange things here is like Deepseek has just gone off and started creating a HTML page, which doesn't really make sense cuz we only want an optimized article. I know why. It's because on the prompt it says creating HTML at the end. That's okay. All right. We'll see what the performance comes back with in a second. For sure, Gemini, even though it's a thinking model, this is way faster than Deep Seek. Look how much quicker it is responding versus Deep Seek. This already finished. Deepseek is still going. All right. So, we're going to test this out now. We'll grab that content, the HTML, and we'll plug it into the preview over here, and we'll see which one performs the best. All right. So, if we compare the titles, so we got the Gemini 2. 5 Pro output over here, and we have the output from Deep Sea version 3 on the right hand side. Okay. So, if we compare the output, first thing that I would say is that the title is pretty bad from Gemini Pro, right? So, it said real SEO training in New York. Stop guessing, start ranking. Pretty average to be honest with you. SEO training New York, the fast track to dominate search. Not bad at all. Honestly, Claude can write much better headlines than both of these, but you get the point. And then what we're going to do from here is we'll compare them side by side. So, if we look at the first line, Deepseek just feels a lot more humanized in his content writing. So, for example, we see here it says, "Need SEO training New York that actually works. Not just theory, but battle tested strategies from someone who's built a seven figure agency. " like that is not bad at all as that's better than most humans could do as an introduction. If we look at this, it's like looking for effective SEO training. Effective being like a weird rule because your website's buried somewhere on page 10. Okay. Tired of SEO gurus selling you fluff. Feeling overwhelmed by conflicting advice. Yeah, I get it. I just don't think that's as convincing versus deep seats output right here. And then I really like this bit. Says here's the truth. Most SEO courses waste your time. Mine won't. and it's all personalized to me and what I sell and it's already selling me from paragraph one. So I would say the output from Deep Seek version 3 actually wins in terms of writing content humanization. I would say both are not as good as the typical content that I get back from Gemini 2. 5 Pro but you get the point. So in terms of coding and this makes sense. building stuff out, Gemini 2. 5 Pro definitely seems a lot better, right? in terms of writing content and humanizing it. Deepseek version 3 is outperforming both. All

right. So, let's come on to another example. And this time I'm going to say there's a tree on the other side of the river and I need to pick it apples in winter. Okay. And then we'll say, but the river is frozen. What do you recommend? All right. And what we're looking for here is it's like a reasoning challenge. So, we're trying to figure out, okay, can it come up with solutions? Can it figure out that the trees actually don't have apples in winter and that sort of thing? But we'll see what it comes back with and see which one. Now, honestly, Google Gemini 2. 5 Pro should smash this. All right. So, you see here how the first thing the Gemini 2. 5 Pro recommends is like assess the apples first, mate. Are they like even worth it? Cuz most apples ripen in late summer and winter. Apples still in the tree in winter are likely frozen, rotten, overripe, or damaged. All right, so first things first, like why would you even be picking apples in winter, mate? That's what it's saying. Whereas, for example, if we go over to Deep Seek version 3, he just doesn't figure out. So, it's like, let's check the ice thickness. Let's get over the river. It doesn't really recognize that actually in winter there are no apples on the tree. So you can see the difference here. This is certainly a lot better for reasoning, a lot smarter also for speed here. It's a lot faster, right? It's a lot faster when we get the outputs. So let's keep going right now. So in terms of just to recap here, in terms of coding, Gemini 2. 5 Pro wins, right? For reasoning, Google wins. And for writing content, Deep Seek version 3 wins. So now what we're going to do is we are going to test this out for the API. All right.

Now the way that we're going to do this is I'm going to go over to open router keys and I've created a API key from open router. All right. And then what we can do inside Visual Studio Code is we can say to root code inside the settings use Gemini 2. 5 Pro experimental for now and we'll say create the best possible SEO ROI calculator you can both for UI and functionality. All right. And that's going to use the API key from Gemini 2. 5 Pro as you can see. And you can see here that actually the API request has failed. So even though this key is free to use inside open router, it's actually not working inside client when we test it out. Now if we switch to the settings and we'll test if deep seats works. All right. So we'll switch over to DeepSeek and then we're going to use V324 which is the latest update and we're going give it exactly the same prompt. Start new chat. Same prompt. We're using DeepSeek here. We'll see if this works. Here's a moment of truth. Which API key works? Taking its sweet time. Yeah. So, you can see here it's finally getting back to us. It's super slow to be honest with you. If you're using the free version of DeepSeek, super slow. Look how slow that is to start creating stuff. Now, let's test out versus Google. All right. So, we're going to go over to select Google Gemini inside the API key section here. Just going to cancel this. So, we'll cancel that task. We'll go over to the API key and then what I'm going to do is select Google Gemini. I'll grab an API key and make sure that we can use 2. 5 Pro Experimental. Is that available here? Actually can't see it. Yeah, we can't use 2. 5. And let's have a look inside R code. Can we use 2. 5 yet for Google Gemini? No, we cannot, my friends. All right, that settles it. So, DeepS might be slow on the API key, but it actually responds. All right. Actually works. Whereas if you're using 2. 5 via Open Router as an API key in client or Visual Studio Code, you can't use it. Right? If you're going to be coding with these tools, then I would recommend Deepseek V30324 simply because it actually works. Whereas Google's API is not available inside client or root code yet. And also with open router, the API call requests just don't work yet. So that gives you an idea. One thing I want to check as well is if we compare these side by side, it's not available yet, but I think that will come soon, which is Deepseek V3. 1. Right? So, for example, Deepseek's models are open source, which means that you can host them locally. So, for example, the last version of Deepseek V3 is available to host locally with Olama. You can't host V3. 1 yet, but I think that is coming. So give it a bit of time and that'll be available I reckon to download and start using straight away as well. Now again just to be clear here you can use you can select Gemini 2. 5 Pro on open router as the API key but it's not working when you call it. So if you go into root code here we'll select Gemini 2. 5. All right save or you're done. We'll give it the same prompt. Start a new chat. When I checked it, it was not working. All right. So, let's just double check. Actually, it seems to be coming back and working now. Fair play to it. Oh, yeah. So, I suppose maybe it's a little bit temperamental, but that seems to work inside root code to be fair. So, actually, it seems like open router and root code work really well, but client didn't seem to work right there. So if you're going to code, Gemini 2. 5 Pro works, but use it inside root code. Don't use it inside client. And then you can get the API keys. And I think that'll be much better for creating content and coding. It seems to be a lot faster than DC version 3's API key as well. And again, it is free to use as well. So let's see what it comes back with now in a second. Wait for that code to load. But then it runs out. Look at that. It runs out right there. Yeah, look at that. So, we actually just building a simple SEO RAI calculator. The API request gets exhausted. And also, for some reason, it says it's using Gemini 2. 0 Pro Experimental, but in the settings, we've selected 2. 5 Pro Experimental. So, that doesn't make sense either. Just something to be aware of. It seemed to get itself back on track there, but it seems super buggy, quite slow, quite buggy. I would be careful using that for now. If you want something super reliable, just use the older versions of Google. you'll probably get better outputs and better response rates. All right, so thanks

very much for watching. If you want to get access to all the prompts, tips, workflows, my best Asian workflows, etc., you can get that inside the AI profit boardroom link in the comments description. Prices are going up at the end of the month, so make sure you sign up now. This is a community, this is a community dedicated to making more money and saving time with AI. And also inside the community, you can post any questions you have and ask 543 members for help. You can also jump on the weekly Q& As's ask any questions you have. And inside the classroom here, we have a crash course on all my best AI systems for saving time, making money, along with all of these SPs, including, for example, a 3hour course on DeepSeek, which you can see right here. All my best AI agents and Deepseek automations. We've got all sorts of stuff inside here, including MCP agents, root code, computer use, etc. So, if you want to get the latest cutting edge stuff, all of the best stuff is inside the AI profit wardrobe. And if you want to get a free onetoone SEO strategy session, then feel free to get that. We'll show you how we take websites from zero to 145,000 visits this month and generate hundreds of thousands of dollars in sales and multiply on this free link building acceleration session. You'll get a custom tailored game plan. You'll be able to ask any questions you have onetoone. You'll learn how to 10x SEO traffic based on what's working for us and our happy clients like you can see right here. Feel free to get that link in the comments and the description.