GPT 4.5 VS Claude 3.7 VS Grok vs DeepSeek: Who Wins?

chat PT 4. 5 versus Claude versus deep seek versus Gro who's going to win we're going to put them all to the test today we're going to be doing some practical examples first off I'm going to show you

exactly what the benchmarks look like how they compare what the API cost Etc and it's going to be very interesting to test these models because a lot of people have been underwhelmed with 4. 5 and its performance and so we're going to compare these side by side now if you're wondering okay how did perform I'm going to show you GPT 4. 5 The Good the Bad and the Moola all right and I've pulled out the special headphones for this so it's a good day that's when you know all right so first of all what you can expect of GPT 4. 5 a lot of people are not sure what's the difference how does it work etc so really this is about like having a better chat more natural conversations less robotic responses better emotional intelligence so it promps is for example to understand how you feel and also react better for your Stakes so less hallucinations you can see that has dropped on average from 61. 8% to 37. 1% and a bigger PR apparently I didn't know out of brain but apparently it does and then what it's not great at and what we probably won't be testing so much today is like maths logic and reasoning and of course it is super expensive stuff if you don't know already this costs 15 times more than GPT 40 so GPT 4. 5 is a lot more expensive now in terms of the benchmark

marks for example if you wanted to compare it say versus the latest version of Claude you can see the GPT 4. 5 is better at chats and facts whereas Cloe 3. 7 is pretty much good at everything when I've tested it GPT 40 is just like generally do everything and you can see the max token limits here the mistakes so this is a main thing is like it has a lot more mistakes the math score isn't great so if you compare it versus o03 mini you can see the math score is way lower and the cost is absolutely redonkulous mate so you probably wouldn't build many workflows with this using like a gentic workflows you're not going to be using this or plugging it into say na10 that would be very expensive and a bit silly whereas you're probably better off just going back and forth with it in the chat if you're not sure already all right so GPT 4. 5 really good a chatting knowledge but O3 is going to beat it hands down problem solving and GPT 40 is way cheaper so we're going to be testing out today showing you exactly what it works how it goes SE what it be really good at as a customer sport apparently writing content we'll test that knowledge based apps I'm not sure but we'll test it and also like vague instructions right what it's not going to be great for is scientific pure coding Etc so let's run through the models first of all one thing you have to know is like if we're comparing Gro here grock actually doesn't have 03 released to everyone right you can request Early Access to Gro 3 for apis but it's not available to everyone me and I post content all the time whereas if we for example look at claw 3. 7 Sonic you can actually get access to that it's pretty good as you can see and also you've got the API pricing for 4. 5 which is a lot more expensive as you can see $75 per million tokens so you know if you're creating a workflow with big outputs like writing or script writing or video transcripts Etc it's not going to be ideal and it's also quite a bit slower I've tested as well especially in beta so let's put these to the test now and I've got a bunch of prompts that we're going to test out the first test that we're going

to run is a social media promo so I'm going to be running this through all models today we're going to be running it through GPT 4. 5 clae deep seek and we'll stick with deep seek version 3 simply because we don't really need to use the thinky version GPT 4. 5 isn't a thinky mode and then we'll just take a transcript from the video so let's grab the latest one right here we'll take this transcript and we'll plug it into our prompts on each and see which one comes best with the outputs all right so we'll take a video transcript and the goal for this is just to create like an interesting social media teaser right if GPT 4. 5 is supposed to be the best at writing content and being creative then let's see how it actually performs versus each all right so we plugged that in and you can see already like GPT 4. 5 pretty slow to respond it's not really going to show you thinking or anything like that it's just going to take a while to respond Claude has already come back with it Goods so deep seek and so has Gro I believe yeah there we go right so let's compare what we got here B DOI just went crazy build apps games and tools and seconds claw three point Sonic makes coding silly easy so it still does make mistakes all right because for example you can see here like claw 3. 7 was smart enough to figure out okay this is not called claw 3. 7 Sonic look at the responses these don't blow me away they're okay you can tell it's obviously AI because no one else in the world is using two emojis at a time on every single line right that's a bit too much if we have a look at the responses here this feels a lot more human so pretty cool attention grabbing teaser right there this is Claude 3. 7 there's not an emoji on every single line and honestly I'm loving claw 3. 7 at the minute for the responses I think that's done a much better job versus chap GPT 4. 5 so for example if we look at this it's am I going to post it like that no I'm going to have to edit it remove some emojis it doesn't read that interesting it's not that attention grabb in whereas if we compare the response from claw 3. 7 which I've pasted below you can see it's much Superior so let's have a look now the way that this is formatted I don't so it's super long lines and it's not separated the responses so that's great and then grock probably coming in second there so if we have a look here it's like wower B dii just teamed up with claw 3. 7 son it make apps and games free and easy boom so your gig while codeing check it out Link in the comments if we have a look fors this GPT 4. 5 I would go with this rather than these again these Emoji are super annoying I do the short lines but even like the formatting right just add a line in between how simple is that if you're really smart as an AI they put a line in between each response cuz otherwise it's going to look weird when you copy and paste that onto Facebook or onto school or wherever you're posting this right onto Twitter Etc so let's just recap here Claude 3. 7 is winning Gro came in second GPD 4. 5 not great and then deep seek I wouldn't use any of these right it's not even formatted it's spelled Sonic wrong doesn't make any sense at least Gro 3 as well was easy could easily figure out okay it's about Claude 3. 7 Sonic which is a sign of intelligence in itself right can it figure out what you're talking about in the transcript without you having to literally spell it out that is a big thing that's going to save me time so I think it's pretty clear who won there and who lost and yeah I don't think there's any more to say on that let's try for email copyrighting now

again we're going to be doing a couple of writing tests just because GPD 4. 5 is supposed to be the go for right in so let's test them out step by step we're going to run this free claw we'll do the same inside grock GPT 4. 5 and their deep seek version three what's amazing about deep SE is like how much it's aged in the space of two to three weeks it just seems like it was amazing when it first came out and then it fell behind so quickly so let's see what we got here I do formatting of CLA 3. 7 Sonic reads pretty nicely hey there ever tried to explain your code impr prom to a rubber duck only to wish the duck would actually write the code for you I've got something better than a coding water foul this doesn't make any sense it's absolutely terrible it's terrible he just un loss for words I don't know what to say it's hilarious testing these out and seeing them perform so badly who is explaining coding problems to rubber ducks and if you can relate to that I don't know what world you're living in it just it's complete nonsense he's after to nonsense all right I hope Gro 3 has come back with something better here so every wonder if you can code like a pro without selling your so to privacy tools buck up because my latest video is about to blow your mind and save your wallet not bad let's have a look what we got back from GPT 4. 5 ever watch your cursor blink for 3 hours waiting for code to actually write itself I've never had that problem with K ey it's not an intelligence response and it doesn't relate to the transcript either so not great there I don't like the response let's have a look what deep seek version 3 came back with ever feel like coding tools either too expensive too complicated or just too much we'll buck up something that'll make you day I like that I think it's got nice placeholders as well it's nicely formatted great hook it actually makes sense it's not about rubber Dock and talking about coding to rubber dock the way that I see this is like Claude is trying to be funny but it doesn't have any context on what funny is you see what I mean and probably if he gave it some trading on how to do that better it would give you better outputs but yeah and the same for this like you just can't relate all right so if I have to rate these step byep I'm going to go with grock as number one it made sense great stuff and also good hook good bullet points nicely formatted Etc I also like the last bit the DIY AI guy who's probably over caffeinated yep that sounds about right and then deep seek version 3 pretty nice not bad at all this bit doesn't really make that much sense so I think you would have to edit it before you send it claw 3. 7 Sonic terrible if probably came out last and then GPT 4. 5 came in third but I just wouldn't use this content like it's just total trash like I've never seen cursor do that all right next one now

we're going to do a reasoning challenge so this prompt right here which is there is a tree on the other side of the river in Winter how can I pick an apple we'll run them through each to see which one performs the best and what we're looking for here by the way is l iCal reasoning and creativity in problem solving the creativity is in can you give us Solutions are you just going to give us problems what are you back so we'll run these now new chat on deep seek 4. 5 and then Claude as well also one thing to note here is like GPT 4. 5 is paid Gro is available for free deep seek version 3 and Claude is available for free up to a certain limit all right so GPT 4. 5 is it's expensive whether you're using the API or not so let's see what we got here given that it's winter this presents a few challenges trees don't bear fruit in Winter apple season is usually laid some fall so it recognizes a problem but it gives us Solutions as well not bad at all deep seek peeking an apple while on the other side of river and which presents a few challenges gives it Solutions but it doesn't recognize right that trees don't have apples in the winter so if we have a look here talks about crossing the river you know how to do it Etc but it's not figured out that actually apple trees in Winter are not going to have apples so we'll put that last so far chat GPT 4. 5 has given us a good idea so it's if it's winter the apple tree doesn't have apples apple trees bear fruit in Autumn let's say there's still one apple left here's what you can do and it feels quite human that's a very humanized response so pretty useful response there it feels humanized reads nicely get straight to the point but it gives us Solutions as well same M Gro Claude has just failed that attempt I want to make it realistic right so this stuff can happen all the time and if it happens throughing the test then I'm going to mark that as a fail because if I have to go back and forth with it or if I struggle with this problem for 30 minutes then it's cost me more time than it's actually worth using the tool so I'm going to put Claude last on because it failed to come back with a response everything else gave us a good response GPT 4. 5 and Gro came in I would say joint first there deep seek didn't recognize the problem of apple tree so it came in third and Claude totally failed and didn't give us the response so it's coming in last next up we're

going to say create an AI powered audit tool for Goldie agency that analyzes the business operations and suggests automation opportunities in HTML so we're going to take that plug this into each now again chat GPT 4. 5 not great for logic and reasoning the maths but we'll test out it will give us a c so it's going to give us something that we can test out Gro pretty good we're not going to be using the think mode cuz I think that would be unfair if we compare it versus non-thinking models deep seek 2 Let's test this out one of the good things about deep seek version 3 and clae as well is that it gives us a canva same with chat gbt 4. 5 whereas Gro you're going to have to test this out in a separate window which is what we're going to do with live weave. com if you ever want to test out HTML Etc so let's see what we get back from each GPT 4. 5 still coding oh we can actually preview look at that I've not seen that before in grock amazing all right let's test this out I never seen the preview before not bad let's test if it works I'm just going to feel this in we'll put GPT 4. 5 analyze my business doesn't actually work though when we test it all out that is not working so it's a bit of a fail from Gro there let's see what we got back from GP 4. 5 we'll allow these and then we'll test out so I'm going to say go the agencies my industry it's got less Fields let's T out industry equals AI pain points equals Automation and time generate the automation plan nothing happens there look at that terrible both of those have failed now we got deep seek has done the same and it actually gives us an automation plan look at that deep seek version 3 has won this so far it's actually working tool right it's the tool is actually working it's actually useful it gives a response the form works okay good stuff so GPT 4. 5 doesn't work Gro the tool didn't work so they're both coming in joint last this has failed again so I'm just going to say retry it now oh my goodness Claud is not working we're going to test it one more I'm going to give it one more chance here and see what we get back here so it's building out the tool I do like the speed of claw 3. 7 as well super useful I'm getting all the names mixed up here it's too many AIS to switch between also I like the canas section and generally when I've designed stuff using claw 3. 7 son it looks pretty nice on the front end as well as the back end but we'll test out it's got to meet deep seek version 3's standard and I'm still Blown Away the Deep SE version 3's tool has beaten everyone else so far the only thing I would say to criticize it a little bit is the design is super basic but you could go back and forth in the chat and ask it to improve that so let's see what we get back now from Claude so we've got the tool back from Claude like you can see which is great the design is super nice right fair play to them but let's test if it works that's the main thing there's no point in creating a tool that doesn't work so we're going to say AI Automation and saving time we'll put in here gp2 4. 5 and look at that it's not working tough one all right so here's what I'm going to say deep SE version 3 is actually one because it just creates it all that works that's what we want get the basics right my friends before we do something Advanced plot 3. 7 Sonic comes in second purely because the front end is better but honestly the tool doesn't work not very useful right GPT 4. 5 totally failed doesn't work at all same with Gro so these two come in last right GPT 4. 5 and Gro com in last for that so who's the

champ we had four contenders just to review so 4. 5 claw 3. 7 deep seek version 3 and grock in terms of the head to head tests Claude won the social media test grock won the email cooperating test for logic and reasoning GPT 4. 5 and Gro one with Claude completely failing to answer and for the AI powered business audit tool deep seek version three actually won while it's cheapy 4 five and grock totally failed with no working outputs so overall rankings we've got the winners and the losers like you can see and in terms of the verdicts Claude 3. 7 one for writing the chat deep seek one for coding jeepy 4. 5 one for General AI stuff and then Gro one as the best free option but it is inconsistent so final fource here GPT 4. 5 is overpriced and overrated all right let's call this Spade So it's 15 times more expensive than GPT 40 but it's not 15 times better Claude is an absolute writing Beast if you need high quality content probably go to Claude deep seek is actually pretty good at coding surprisingly and Gro is fun but pretty inconsistent probably still not on the same level as the rest honestly which AI do you think is winning let me know in the comments so thanks so much for

watching if you want to get access to my best prompts tips SS and workflows plus AI agents feel free to get that inside the AI profit boardroom we have a beta price on right now so if you join now before the 10th of March you'll save 20% the prices are going up in just a few days and if you sign up now you lock in your legacy price if you sign up later obviously you're going to miss out inside the classroom you're going to get all of my best trainings on email Automation and content automation social media AI agents video Automation and how I build AI avatars AIC automation Q& A call recordings all of my best saps there's an insane amount of saps like you can see inside here all of it gets updated every single day every time I do a new YouTube video I put a new sap inside the AI profit boardroom and on top of that it comes with an awesome crash course and all these cool stuff that you can automate and it's a community of 233 people so you can post in the community ask any questions you have and get help whenever you need it like you can see every single post all these posts get replies and answers and if you want to jump on Call Of Me each week you can jump on the weekly q& as jump on ask me any questions happy delpia feel free to get that link in the comments and description to the AI profit boardroom top of that if you want

to get a free SEO strategy session that shows you how we take websites from 0 to 145,000 bit this 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're going to get a fre SEO domination plans this is a custom tailored link building plan so you can generate more lead sales and profits from your website you just go the secret SEO link building where answer any questions you have you learn the best link building strategy for your website plus add with competitors building and how to 10o traffic based on what's working for us and a happy clients like you can see right here feel free to get that link in the comments description appreciate you watching thanks so much for watching byebye