NEW DeepSeek V3 0324: DESTROYS Claude 3.7? 🤯

Deepseek version 3 versus Claude 3. 7 Sonic. Who wins today? We're going to test this out and see who's the winner. Which one is the best? Now, if you're

not familiar with this already, just a few hours ago, Deepseek released V3324. Now, this is available inside the chat. So, you can go inside Deepseek. com, chat. deseek. com, and get free access right there. or you can go to LM Arena, go to the direct chat and you can select DeepS version 324 there as well. And additionally, if you prefer, you can also grab the API key for free on openroo. ai or just go to the chat room. So it's openrooout. ai/hat and then you can compare two different models. So you can see here we've got 3. 7 Sonet versus Deepseek version 3. So, what we're going to do here is we're going to test them side by side. See, okay, who's the best? Who's performing the best? Which one creates the best prompts. Now, the first

thing that I'm going to test is its writing ability. One thing to note here is the 3. 7 Sonic is a reasoning model. It has thinking abilities, whereas, for example, Deep Seek version 30324 is a base model, right? So, it's not a reasoning model. It doesn't have thinking capabilities. It should be quite smart, but at the same time, I don't think it's going to be as good at reasoning. All right, but we're going to test this prompt out now. So, I'm going to say create an SEO optimized article for this. We'll put in the keyword and then we'll plug this into both. Right, so you can see Deep Seek working its magic here and Claw 3. 7 working its magic as well. Now, if we actually compare these side by side, what's interesting about this is actually it looks like Deep Seek version 3 is totally given up on the response set. I'm just going to refresh that. So, it seems to be stuck on processing. That is super buggy. What I might have to do is just open up a separate tab. So, Deep Seat version 3 is now writing as well. We'll open up both of these. So this is the output from Claude 3. 7 set. One of the first things you're going to notice here is you see how it says 2023 inside the response here. That is not ideal my friends. We don't want that. So claw 3. 7 seems to be a little bit confused there already. Deepseek version 3 has come back to us. It's frontloaded the keyword. It's included in the first line etc which is great. Let's have a look at the content from Claude. Also, look at this. Right. So, Claude 3. 7 Sonnet said, "Looking for top-notch SEO training in Manchester. " Now, the key word here is top hyphen notch. So, if you actually look at the outputs here, number one, claw 3. 7 Sonet is put 2023 for some reason, and number two is using AI fluff words like topnotch. Whereas, if you compare that side by side versus Deep Seek, this feels a lot more humanized, right? So if you look at the first line, looking for SEO training in Manchester that actually works like no AI fluff feels quite humanized. Gets straight to the point. It's linked to our funnel after the first paragraph. All right. So if we pull these up side by side, it looks like Deep See is actually doing a better job than Claw 3. 7 Sonet when it comes to writing the content. I'm just going to check how long this is. So, the content from Deep Seek V3 is 485 words and the content from Claude is 1,087 words. All right, so let's just pull these up. What I'm going to do is

run them through Zero GPT and AI detection tool and we could see which one creates the most humanized content as well. Honestly, if I had to pick both blogs, I'm going to go with Deep Seeks because it just feels a lot more humanized, natural, etc. So, let's take this content and we'll run it through zero GBT and just see how it performs in terms of humanization. So, Deep Seek, this is pretty amazing actually. Deep Seek's content comes out at 0% AI detectable, right? So, the AI cannot detect whether this is human written or not. It thinks it is actually human written, right? So, Deep Seek has done pretty well there. Now, let's compare 3. 7 Sonet. So, we're going to take this content and just compare it side by side. Bearing in mind, Deep Seat scored 0% which is perfect. Claude 3. 7 has come out at 0%. Human sector as well. So, they've done they've both done really well in terms of writing the content. I would just say that Deep Seeks doesn't read quite as nicely as using Sorry, I wouldn't say Claude's content reads as nicely as Deep Seeks. All right, next up, what I'm going to do is I actually prefer using these directly in the chat. I think it'll be easier to compare as well. So, I'm going to get rid of Claude right there or XL that. And we're just going to use Deep Seek inside Open Router. And then we'll use Claude over here. All right. So, we'll run the next prompt. We're going

to use a reasoning challenge to compare them. So, I'm going to use this prompt here, which says, "There's a tree on the other side of a river in winter. How can I pick it up? " Basically, what we're looking for is like solutions. We're looking for the AI to recognize that obviously apples don't grow in winter as well and that you could be wasting your time, etc. So, let's see how it performs and whether it can actually figure that out. All right. Now, you see claw 3. 7 is actually a lot faster. So, this is still loading with Deep Seek over here, but with Claw 3. 7, it's already given us the answer and gotten straight to it. So, in terms of speed, Claw 3. 7 is winning. However, this model has just been released, so it is going to be a little bit slower than normal. I'm waiting for this to load, and you can see it's taking quite a long time. I might switch to the paid model if this doesn't work. But so far, it's letting us down. Let's just go back onto Deep Seat V3 and we'll use it directly inside the chat. See if that works instead. So, we're going to use the same prompt inside the reasoning model instead. Here we go. So it's analyzing the situation and you can see it's actually got the reasoning ability here. So it's like right apple availability. Apples are generally harvested in late summer to autumn. So by winter most apples have fallen or been picked unless it's an evergreen or a special variety that retains fruit which is uncommon, right? So it's figured out. Okay, there's a lot of problems with this. It's also really in depth in its analysis. Look at this. like understands the scenario, analyzes the possibility, assesses the river potential solutions as well. So, re-evaluates the initial assumptions as well. It does feel like a reasoning model even if it's not. If you compare the output of claw 3. 7 Sonet versus Deepseek, genuinely Deep Seek is thinking this out a lot more. You're getting a better output from Deep Seek for sure when it comes to reasoning. So I'm going to say that Deep Seek won this reasoning challenge simply because it gave us more solutions. Analyze the situation a bit more and also it was a lot more focus on analyzing the situation than coming up with solutions, right? Whereas for example with Claude is really short with its response. So, if we take a look here, it's like this is a peculiar situation since apples don't bear fruit in winter, which is great, but it's not really analyzed the situation in as much detail as Deep Seek. If you need to really reason something out, this is going to give you the most logical and best answer. So, Deep Seek wins on that one as well. Now, let's compare on another challenge. What we're going to do here

is we're going to test the search model. So this is web search using deepsek. Now if you're outside the US, I'm just going to double check the settings here, but if you're outside the US, you won't get the option for web search. So Deep Seek actually has web search, but I want to make sure it actually works. So I'm going to say, okay, what happened in AI news today? So just to be clear here, Claude only has web search for people in the US. If you're outside the US, you won't get that. However, the search service inside Deepseek is not working either. You can see it says due to technical issues, the search service is temporarily unavailable. So, both of these failed when it comes to that. Now, you could go inside open router and select web search here and use the same prompt. Right? So, if we go inside open router and grab that, open a new room, web search selected, but you can see it still fails from what I can see. So, we've selected web search, right? We've got that selected on and I've said what happens in AI news today and it still thinks it's June 2024. So, not ideal there. I would say in some ways it's better that Claude just tells you we don't have it, mate. Rather than wasting your time trying to use a search that she doesn't work inside Deepseek. So, I'm going to say Claude one there, but both of them don't really do that. The other thing to note here is, and I don't know if this is going to be available to host offline yet. I'm sure it's coming to O Lama, but yeah, this was only updated two months ago. So, the new version is not here. You can get it through hugging face. But the main thing to note here is that when you're using Deepseek, it's an open source model. So, if model, you can host it locally. I'm sure it will come to OAMA pretty soon cuz Deepseek version 3 is already on there. The other option that you have is you have a free API key, right? So, for example, if you're using Claude, you get charged for tokens. It's pretty expensive, right? If you're using the model for API calls, whereas for example with DeepSeek, you can go to open router, you can type in DeepSseek over here, select 0324 for free, and then you can actually grab a free API key from this and use that inside open router for coding, right? So you can build like AI agents or you can build automations or you can do whatever you want for free using deepseat version 3 which kind of beats claw 3. 7 and sonic as you have to pay for that and a lot of people can't afford it right that's one of the most common sort of problems I see on my YouTube comments is like people complaining that they have to pay for this stuff so you get the point like this is open source and number two this is available to use inside client ru code open router chat room etc and it's available to use for free as well. So side by side, I would say Deep Seeks version 3 is winning now for content creation. It's winning for reasoning. It's winning for the free API key. It lost on the web search test. What else can we test here? So the next thing that

I want to test is actually coding, right? So if we go back to the chat room here, actually coded this earlier. Where is it now? So, I ask it to create a beautiful landing page for my SEO agency, right? It coded this HTML code right here. So, if we preview this inside liveweave. com and plug that in and it actually created a really nice page, right? Really nice design. Simple, but it does the job. It looks nice. It's clean. It even came up with the brand name, etc. Now, let's compare this versus Claude 3. 7 Sonnet. So, I'm going to give exactly the same prompt. just keep it basic, which is create a beautiful landing page from SEO agency. Where's the chat here? We'll plug that in and we'll see how it goes. Now, some people are saying that there's only improvements in maths and coding when it comes to Deepseek. That's an interesting one. I've not seen they've not actually released like the official benchmarks for this new version of Deep Seek, so it's going to be hard to compare, but I've definitely noticed a difference in reasoning. Right? If you compare the old version of Deep Seek version 3, it wasn't very good at reasoning. it did fail in a lot of recent tasks. Whereas, for example, when we looked at that crossing the river riddle before, DeepC version 3 actually outperformed core 3. 7 Sonet. So, it's really interesting to see that. So, let's see what we get back here. Just waiting for that to load. The other thing that we can test in a separate tab is creating a cool 3JS game. Right. So, if we have a look at that game that we created earlier, this one we take a look. So, we've got the output from Claude over here. Let's take a look. Actually, very similar, isn't it? Look at that. I'm just going to publish this so that we can preview it full screen. So, let's open this up. So, if we have a look through this, I really don't like the way Claude doesn't ask you for an image, but it leaves a little blank space. I never like that cuz it makes the page look a little bit messy. So this is using claw 3. 7 on it as you can see and the page itself is really thin. So all it's got here is a little title bar, a little image and then that's it, right? It doesn't even fill out that bit properly. Whereas for example, if we go to the deepseek version actually created the full page, designed it nicely, added a footer which the output from claude doesn't have. Additionally, you can see that we have much nicer title, headings, etc., nice icons. Like the output from DeepSeek is ready to go. Whereas the output from Claude is really poor, right? You wouldn't use that. You can't publish that. It's not ready to go. Now, it does say it hit the max length for a message. So, I guess we could say keep going, but then you're wasting more time. You're going back and forth with it a bit more. It's going to be a little bit more frustrating to use. Next up, we're going to test another coding challenge. So, I'm going to say inside here, create a sleek conversion focused landing page for Goldie agency's automation dashboard. The design should be modern, etc. Here's what to do. So, we'll plug that in and then we'll do the same inside Claude. Open up a new chat and see what we get back. That is not ideal here. I was asked it to create a sleek modern landing page and it's not actually recognized that it needs to code something out, right? So, Deepseek has failed on this one right here. Whereas if we go over to Claude, I'm 99% sure it's like smart enough to figure out, okay, this is a coding task, right? It needs to go off and code. And that's exactly what it's done right here. So Claude is already winning by figuring out, okay, this is a user intent of what we want. Whereas Deepseek is feels a bit messy. It doesn't understand what we want. It doesn't have that intelligence to figure out, okay, this is what we're talking about, etc.

But yeah, I think overall if you're looking for something that's free to use, gives you good outputs, pretty good at coding, good at reasoning, good at writing content, then I would recommend that you go with Deep Seek. But if you are looking for the best of the best, the best UI, the best outputs, the highest quality content, a smarter, more intelligent model that also comes with reasoning, then I would recommend that you stick with Claude. I don't think it's going to be clawed anytime soon. So, thanks so much for watching. If you

want to get a free SEO strategy session, feel free to get that link in the comments description. We'll show you how we take websites from zero to 145,000 business month and generate hundreds of thousands of dollars in sales and autopilot. On this free link building acceleration session, you'll get a free SEO domination plan. So, you'll get a custom tailored link building plan so you can generate more leads, sales, and profits from your website. and we'll be able to answer any questions you have. Plus, 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. And if you haven't already, check out the AI profit boardroom. This is a community focused on making more money and saving hundreds of hours with AI, right? So, if you want to get my best automations on email and content, social media, AI agents, web automation, ASEL automations, crash courses, and SPS, feel free to get that link in the comments description. Also comes with weekly Q& A. So, you can jump on the weekly Q& A calls with us. Ask us any questions you have. And inside the community, you can post and ask any questions, and then you'll get help from 530 members, all focused on the same goals as you. So feel free to get that link in the comments description.