Baidu’s Ernie VS DeepSeek: Who wins? 👀

Ernie X1 versus Ernie 4. 5 which has just been released from BYU versus deep seek R1 and deep seek version 3 So today we're going to be testing whether the new release from BYU Ernie 4. 5 and X1 actually outperforms deep seek R1 and deep seek version 3 now if you haven't

seen the benchmarks right here you can see the Benchmark comparisons the new models from Earnie 4. 5 and X1 which have just come out from China are actually outperforming on the benchmarks the other versions of deep seek and even chat gbt 4. 5 now the other thing to know here is if you get the API from BYU for Ernie x14. 5 it's actually at 1% of the price of chap cheb 4. 5 but it's outperforming chat gbt 4. 5 on many benchmarks and the same for deep SE car one as well so if you want to have a look at the price comparisons on the apis and you can get free access to this as well inside the chat which I'll show you how to SS 2 in a minute but you can see the model comparisons here so for input look at the price comparison $75 per million for input on GPT 4. 5 whereas ear is 0. 55 deep C car1 0. 55 and if you compare the models which is nnie X1 is a deep reasoning model you can see here that it's basically half the price of DC R1 right so deep seek R1 is 0. 55 per million inputs and Ernie X1 which is basically like by's alternative to deep C Goan is 0. 28 right so you can see how much cheaper it is but it's also performing really well on the benchmarks so this is going to be very interesting you see the benchmarks here so if we have a look we've got ear 4. 5 versus DC version 3 so these are comparative models along with GPT 40 and GPT 4. 5 and you can see on many benchmarks for example like C eval or for example C maths you can see here the ear 4. 5 is outperforming deep seek and chat GPT in many ways so let's compare them side by side the first thing that I would say

that is a little bit annoying it's not perfect right so for example if you actually go on to deep seek is all in English it's really accessible it's quite easy to sign up for an account with email that is much easier to get access to than BYU actually took me a while because I have to translate the website into English and then also I had to sign up with a phone number that had to be verified right so it's not as easy as using a mobile and additionally when you're using these models you have to figure out okay what does each of these mean and it's just not a Syle also for signing up to the API I don't even know how to do it but I do know how to sign up inside the chat which you can see right here and you can get that available here I'll leave a link inside the AI profit boardroom to the sap and everything like that so if you want to get access to all this just check out the AI profit boardroom but yeah let's compare the models now side by side so if you want to understand okay which is the comparative model right so if you're using these models let me pull this up here and we're going to translate this to English so make sure that in the top right here on Google Chrome you select from Chinese simplified to English if you're wanting to use this in English essentially in the top left you can select between the different models right so if you want to use X1 is basically like deep seek R1 right so these the comparative models this is a reasonable model and the same with DC G one if you want to compare for example by's model for AI versus DC version 3 which is a lot faster but not a reasoning model then you would select 4. 5 so this is Earnie 4. 5 and we can actually compare them side by side versus deep seek as well right so if we go to deep seek. com then we're going to deselect our one and this will allow us to use a comparative model versus deep SE versus ear 4. 5 so this is ear 4. 5 and then over here we have deep Siege version through so let's run a few tests in terms of how this performs side by side which one performs the best Etc

and one of the first things we're going to do is we're going to test this for writing skills so I'm going to compare deep seek versus Ernie for writing okay so what we're going to say here let me grab this prompt and this is one of my favorite prompts for creating SEO content we're going to grab this right here and we're going to go over to Earnie 4. 5 and we'll say create an SEO optimized article for this put SEO train in Japan now one thing to note here as well is like these models are in Chinese when you're using the UI you can translate them like you can see and then additionally when you're comparing them this will give you English outputs you can also choose to have online search selected I think it has a ler chance of working if you do it that way and also here what's interesting is you can actually upload video upload audio upload pictures and upload documents right so you can feed information into this and if you want to go full screen on this prompt you can do that so let's see how they perform we'll start with Earnie 4. 5 and it says something went wrong please regenerate so we'll try that again there we go so we get the output wait let's see what we get back in a second that doesn't seem to be working at all so it's a bit of a fail already so far we'll compare that versus a deep seek version 3 and then we'll compare that versus Ernie X1 and deep seek R1 so let's see what we get back here we're going to try and generate that again that is crazy right so it's not working at all let's try a simpler prompt already deep SE version 3 is winning simply because it works by default but let's see what we get back so we'll try and one here bear in mind like there are going to be a lot of people using this AI model right I'm going to start a new chat here because that is just not working at all right so far Earnie 4. 5 has totally failed deep seek version 3 has beaten it on the challenge let's have a look now at Deep seek R1 versus Ernie X1 hopefully it does better than Earnie 4. 5 the thing to not here is like this just got released 24 hours ago so I guess there's a lot of people using this and that might overload the servers but at the same time you would expect this to work look at that so if we compare the thinking an action process it really doesn't matter what happens on the benchmarks if you compare it in reality and this model is not working properly then you know what's the point right this would waste your time if you came to using it properly so bit of an issue right there what I can do is pull up some examples of what we've created earlier today using these models but if this doesn't work and deep seek R1 does work then by default deep seek is winning right so

far now let's start up a new chat here and what we're going to do is we're going to run through some of the previous examples that I've used Earnie 4. 5 and Earnie x14 right like you can see so what we asked it to do originally was create a self-playing snake game using HTML of simple GUI right so far deep seek 1 and deep seek version 3 have already won the first challenge of writing let's test it again and now what we're going to do is we're going to say create a self-playing snake game using HTML with a simple GUI inside deep seek R1 and we'll do the same inside deep seek version 3 and we'll see what we get back now some people asking like how do you actually build the interactive Games Etc desktop games applications so this is how you do it so you just keep the prompts really simple and you ask it to do that right now if you want another tool for this we've actually got something inside the AI profit boardroom which is custommade tool for building out stuff like this right it's called prompt Forge let me pull this up you so this custom GPT right here is inside the AR profit boardroom you can use this to generate prompts for your games applications Etc and it will generate much better prompts simply because it's more in depth and that sort of thing it's custom designed for that stuff so with we've got the shimo from Ernie X1 over here let's grab that we'll go into live and we'll test it out so this is the response from early X1 pretty good to be fair I've never seen the Snake Game move that fast working pretty nicely right there it doesn't play itself for too long seems to break but it's working it does exactly what we asked it for now this is the output from ear 4. 5 let's test this out as well see if it works for the self-playing snake game and that's working pretty nicely too right you can see that's working it's generating the responses it's going so far but it does stop playing as well and the whole point of this is It's supposed to be self playing able to play itself right now if we actually take the output from Deep seek so we've got the deepsea version three output from snake right here and we'll compare these side by side all right so let's grab that so this is the output that we've got from Deep SE version 3 grab that too and then finally grab the JavaScript plugin there and the CSS over there now that is not really a self-playing snake game right if you look at this is a weird game I wouldn't say that a snake at all it seems like a modification of it and so far if you compare deep C version 3 versus Earnie X1 this is the output from Earnie X1 and 4. 5 and X1 are beating it for coding right so deeps version 3 doesn't seem to perform as well for coding now let's check the output from R1 all right so let's grab these we'll get the HTML back from Deep seek R1 for the self playing snake game we'll see how that performs so click on start game to be fair that so far seems like the best output although it doesn't seem to play itself for too long let's see how it goes so scores three I do the UI as well you can press these buttons and then flick between it yeah so I'm going to say that's the best output so far so deep SE car one is coded the best I would say Ernie X1 is coming second there Ernie 4. 5 is coming third and then deep seek version 3 created some weird version of snake I don't even understand right so for COD in deep C car1 is still the best but only X1 comes in second so fair play to it now what we can do is we can compare side by side versus some other stuff so let's see if this actually works we're going to grab another prompt here and

this is going to be a reasoning challenge right so basically we're going to use a winter Apple puzzle and we say there's a tree on the other side of the river in Winter how can I pick an apple and what we're trying to compare here is the logical reasoning you know does it recognize apples don't grow in winter and also it's creativity in problem solving so does it give us solutions to give us lots of solutions Etc so we're going to grab this prompt like and we'll plug that into Ernie X1 and we'll see if this actually works I don't have high hopes for it but we'll see if this works or not and we'll do the same inside early 4. 5 as well maybe you can't run both models at once maybe that's the issue I'm going to come back to X1 in a minute let's try this is it working or not working yeah so only 4. 5 is just broken that is lost by default let's try X1 now so this is the Deep reasoning model hopefully that does better thinking mode is really weird like the outputs here maybe you would get better outputs in Chinese if you spoke simplified Chinese but we're comparing these things in English right we're comparing these models in English and all the announcements were in English on Twitter as well right like you can see so let's see what we get back this doesn't really seem to make much sense in terms of reasoning it's just like alternative methods seing and it just cuts off mid sentence another idea is what about using a and then it cuts off and then it says something went wrong please regenerate all right total fail okay so let's try now we'll try a deep seek version three and we'll see whether that can solve the job so far X1 and Ernie 4. 5 have totally failed on this hopefully we get better outputs from Deep seek R1 so there we go so it's definitely coming up and figuring out the some of the problems here it doesn't recognize the trees don't grow fruit in winter so it's failed a little bit on the reasoning there but at least it's given us some sort of output we'll compare it versus deep seek R1 now as well and we'll see what we get back here so it's thinking about things still can't believe that X1 and early 4. 5 are just not working maybe when they're actually working they're better and outperform deep seek but so far I think all of the promises that we've got right here in terms of benchmarks just don't seem to be reasonable in reality if the model just doesn't work deep C car1 seems to be struggling here as well so we've got no outputs from Deep one which is a bit of an issue too the other thing that I'm going to say here whilst we're waiting for that to load is the by's models are not open source right they are planning to be open sourced from June the 30th so far yeah deep SE car1 has failed there so the only model that actually worked here so far was deep seek version three so deep seek version 3 comes out on top for the reconing challenge just by default even if it didn't recognize that trees don't have fruit in Winter now let's come on to two other ideas here that make deep seek still a better version right so if you type in deep seek into open Mara you've got a bunch of different apis you can use both distilled via llama and also for free right so for example deep seek i1 Zer and there seems to be like newer and newer models coming out all the time so you can see for example deep seek r10 was created just 11 days ago and you can get a free API from open rout and then you can use that to build your AI agents or tools or applications whatever you want right whereas if you type in Ernie you won't find Earnie on open routa so it's not very accessible and the same on as well so if you want to use deep seat inama so that you can host it locally you can run it locally Etc then you You' got all these different versions of deep seat that you can run locally right which is awesome CU then they can be offline you got more privacy safety you can use this on like AI agents with our o l integration which is awesome right it's completely free as well whereas if for example you type in Ernie you're not going to find Ernie X1 or Ernie 4. 5 on a llama IR it's not available on open routa and a llama so it's not that accessible and that's one of the biggest criticisms I would say of about it as well not only the fact that it's just not really working consistently right now and maybe that's just cuz it's in beta but still not great whereas if for example you're using deep seek is more reliable you can access it more places you can host it locally it's open source Etc that just makes it a more powerful model overall so thanks so much for

watching if you want to get access to my courses on how to use AI how to run it locally how to set up op Manus locally we actually have a full 3-hour course on deep seek that you can see right here along with all of our bests and AI agents with deep SE feel free to get that link in the description the prices are going up for the AI profit Board Room at the end of the month so make sure you sign up now if you'll lock in your legacy price if you don't sign up now then you'll miss out on that and you have to pay more in the future right so if you sign up now you basically lock in a cheaper price and get a lifetime deal on that and additionally if you have any questions feel free to post them inside the AI profit boardroom Community inside here we have 46 4 members ready and willing to help you out so feel free to get that link inside the comments description and also you can jump on weekly q& as with us and ask any questions you have additionally if you want to get a free one to1 SEO strategy session feel free to get that link in the comments and description on this free SEO strategy session we'll show you how we take websites from 0 to 145,000 this month and generate hundreds of thousands of dollars in sales on autopilot based on what's working for us and a happy clients you get a custom tailored link buing plan to gener more leads to sales from your website you will discover the secrets as you link building right any questions you have one to one you'll learn the best link building strategy for your website based on what's working for us and our happy clients like you can see right here feel free to get it Link in the comments description and appreciate it