Why ChatGPT o3 is BETTER than Deepseek

so I want to show you the differences that have been made here so take a look at GPT 40 right GPT 40 had a 2.5 per million tokens cost right that's expensive in comparison to this right so it was literally twice as cheap to use deep seek when comparing to like 40 which is not a reasoning model but then when you compare it to 03 mini now we're price matching hi my name is Demitri panichi and I'm a content creator agency owner and AI Enthusiast you're listening to the AI agents podcast brought to you by jotform and featuring our very own CEO and founder idin tank this is the show where artificial intelligence meets Innovation productivity and the tools shaping the future of work enjoy the show why would you use deep seek now that chat gb03 is out and honestly why would you use

the chat PT Pro Plan these are the questions that we're going to be answering in this video and podcast of the AI agents podcast so as you probably saw a couple weeks ago chat GPT released open ai03 mini and o03 mini high now 03 mini sounds like something similar to what you had with Chad gb1 and also if you go and remember this there was a 40 Mini model now mini was something that you were just used to having it was the smaller version in regard to the computational power and comprehension and thinking capabilities of 40 and 01 however with this new release this chat p03 mini is the newest and most efficient model in the chat PT series when you take a quick look at its competition math scores it did better than 01 and it did significantly better than 01 mini and as

you can see there's 01 mini 01 preview And1 and there's 01 mini low 03 mini medium and 03 mini high and that's exactly what you're going to see inside of here too you have 03 mini and 03 mini high so the O3 mini option at its base is going to be good for fast Advanced reasoning and the o03 mini high is going to be the most logical and in-depth of the options we'll scroll down here to HD level science questions and you'll see how it fares with the rest of the different models here it's the smartest it's the best and it continues to be massively better at coding than it was previously which as a lot of us know this is where the huge gains are going to happen because it's going to Compound on itself when the developers are going to be able to

use the large language models to write better large language models better computational capabilities it's just crazy to me so there's continued Improvement across the board here and I like that they're going to have multiple different reasoning levels so you can have a different amount amount of like execution time and output and requirements are going to be different across different types of tasks so that's totally good with me and what's awesome is we've gotten the latency down significantly as you can see this is a little over 10,000 Ms and this is down to about 7500 Ms so about 2500 milliseconds less which might seem minimal but that's a pretty good jump and the speed at which you experience these things is always going to be important so let's be real all of you are going to be happy with that now who gets this plan for

the first time you actually get a reasoning model available on the free plan so while I do have the team plan it's popping up there not cuz I have the team plan it's popping up there simply because you get it if you are anybody and to me that's nuts and amazing if you are on that free plan you simply are going to select the reason option in the composer or regenerate IR response to try the O3 Mini model so does 03 mini replace 01 mini yes as you'll see inside of of my jgpt account there is no 01 mini anymore and not even in the more model section and that kind of tripped me up during the beginning of the recording of this video and what's really great is right away we're getting the 03 Mini model in the API and just so you know

no O3 mini does not support Vision or image inputs yet but you can use that with 01 which was not the case prior now when it comes to pricing I did want to bring this up especially with deep SE cuz in the last video where we died dived into deep seek it's pretty simple I was like guys the pricing is not even close right like the the 1 million tokens input and output prices were literally like 30X cheaper in uh deep seek so I want to show you the differences that have been made here so take a look at GPT 40 right GPT 40 had a 2.5 per million tokens cost right that's deep seek When comparing to like 40 price matching so if you take a look at the cost of all of these right it's a pretty significant price difference so let's take

a look at the output cost so the output cost was $10 right for a non- reasoning model $219 was what it was costing for a deep reasoning model and a general chat was $110 so as you can see it was like a 9 to 10 cost increase that's not great and I'm not saying $9 to $10 which I guess it is kind of $9 to $10 but that's be 9 to 10x the cost which is not great now what we have is 03 which has significantly reduced the pricing of that to $44.40 which is still twice as much as what deep seek Reasoner is but a 2X increase is a lot different than a 10x increase so the pricing on this is better so you don't really need to worry about deep seek I don't think anymore I think they're going to get the price

done again by the time like the full 03 model uh comes out and they always like continuously decrease the prices on these things as things get more efficient so I would just stick with the model that I trust more which is going to be 03 with factual information and the like deep seek came hot off the presses but when it comes to chat gpt3 I really think that when I take a look at mini high it's going to do a good job so let's just play around and what I like to do to test models out is say what are some prompt examples I could use to test out the abilities of chat gpts 03 mini High Model A lot of times asking Chad GPT what you can ask it ends up being the best answer so to effectively evaluate it so yeah yeah let's

give it something difficult okay ooh I always like the philosophical questions I want to see how it does in a short story as well and a logical puzzle so let's let's try these out okay I'm going to go back and forth between the two so couple things by the way you do have the ability to with O3 Min High add a variety of different things you can add images you can search what's interesting you can add files and what is interesting is inside of this I can give it this sales deck maybe actually and I could say find some ways to integrate a more a value focused pitch deck for Content Services clients that focuses on how educational content reduces the costs of customer service agents and how online content drives growth for sales in SAS so what it's going to do is it's going

to analyze the document this is a much improved thing because you didn't have the search capabilities before now you have search and you have the ability to utilize the the PDF that I added and as you can see it's doing a great job of really breaking it down in depth going through everything I have so for example defining the two key value propositions talks about the problem about how in a crowded sass Market Genera in quality leads and nurturing prospects through sales funnels challenging and the solution is building a robust content engine and then this as well for the customer support section also gives me the problem and solution and I didn't tell it to give me the problem and solution format but I know that's a really great way to Showcase value for sales so definitely want to recommend this for anybody trying to

improve sales decks and that kind of stuff and as you can see it does have a reference to what was in the PDF too now let's get back to asking it some fun questions okay so I want to do a logical puzzle to see how it does you have two ropes of uneven thickness that each take exactly 1 hour to burn how can you measure 45 minutes using these ropes and as you can see it's going to go through the logic right here I don't have to show it as it's going but I can if I want I will say the response time is pretty quick which for the 01 mini and 01 in the first place were pretty slow 01 preview was really slow if you guys recall that so start both ropes simultaneously after 30 minutes repe will have completely burned out because

burning from both ends has its total burn time at the exact moment light the other end of the Rope B in the final 15 minutes in the final 15 minutes rope B now burning from both ends will take 15 more minutes to burn completely in total you've measured 30 minutes from Rope A plus 15 minutes from rope B which adds up to 45 minutes okay I can see how that works out that makes sense so I find it interesting how it's working through logical problems and explaining it for us so let's try another one I really like testing out the capabilities of these things all right let's do some math prove that the sum of the squares of any two sides of a right angled triangle equals the square of the hypotenuse it goes through and shows all that math that quickly that's insane and

it's I feel like the latency is way less like I'm not seeing like a weird lag as the text coming out it's like Smooth they probably added some nice CSS so there's like a a slow fade in of things too it's not just like slapping it onto the screen you guys know what I mean like it used to slap it onto the screen or like kind of swipe it in a harsh Manner and now I feel like it's not someone typing insanely fast it's it's fading onto the screen as it's going in fading in so it knows how to do this deep math now let's go back to one more test that I want to do which is discuss the concept of Free Will in the context of determinism this is a good question um this guy's going to be logical I hope see if

it's going to properly get this one so I was going to go through the whole logic here real quick all right so determin M core idea is determinism holds that every event basically is determined right including human actions it's inevitable nothing can really change it so this is an argument between determinism and free will uh I would fall on the Free Will side because even if we are in a deterministic society like me trying to prove to you the fact that Free Will exists or even you know probably making my argument a little bit more staunchly you can't prove to me that determinism is true by making an argument for determinism in the confines of determinism because you would be forced to believe or not believe in determinism it would kind of be irrespective of what I'm saying and even if you were convinced by

what I was saying you were predetermined to be convinced by what I was saying and everything that led up to my situation as well uh made this happen so the argument here though is that it's more logically consistent because if everything is a part of the natural world and we must conform to these laws then true Randomness wouldn't exist which I can appreciate that uh I just don't believe in uh naturalism so that's why I would morall on this end but this does give a a great breakdown of both sides you can't really have like ethics and responsibility in determinism if you're being logically consistent and this is a a great point that it points out here and it even brings up Quantum indeterminancy which is a major part of that so really cool stuff here definitely a Smart Cookie we showed you the stats

we showed you the way that it functions we showed you that it has all of these capabilities like the temporary chat that obviously every one of the models has but it has uploading files photos searching the web taking screenshots all of these things are minor improvements that they're rolling out over all the models but aren't necessarily capable or weren't necessarily capable for the 01 preview in the beginning and even in 01 still you can't search the internet which is very intriguing to me that that's the case but 01 mini high you can still se you can search the internet and it's allegedly a better reasoning model and I will never know why they can't do that or maybe not never know I will and I don't know why you can't use the search in that capability but I'm sure it'll get there what I'm really

excited about to see is when those different capabilities are available not only just for us on the desktop app but also ideally in the apis we can have images and that kind of stuff be supported these reasoning models are going to give you long-winded responses so if you're looking for more direct responses and you're looking for more day-to-day tasks you still going to want to use the 40 model but this kind of makes me think that you don't need chat GPT Pro Plan like I appreciate chat GPT I use the team plan right which is $30 per user per month I do Bill it annually so it's a little bit better but this $200 a month I just I really don't see the justification at all anymore especially when you have 03 mini on the free plan like a very high-end model here you have

01 on here which is going to get upgraded I'm sure um you have the pro model I guess here you're going to have more Sora video Generations you have access to operator uh we've talked about that before on this channel you have unlimited access to Advanced voice with higher limits for video and screen sharing I just don't see the justification for the higher amount when you have 03 mini high on the free plan and you're getting more of it on the Plus or the team plan and you're going to have more than enough use so the fact that you get on the plus plan some good things and then if you get the team plan you could seriously just pay for the team plan and only have three people which will set you back 90 a month and then you get also the higher messaging

limits and you get access to the models with file uploads data analysis canvas projects tasks and search that you don't get all the time in the plus plan across everything so it's like I don't know why you'd go for the Pro Plan I really don't see it like you go for the plus plan or you go a little bit higher for the team plan if you want a little a little bit more juice on specific specific things but even that's like less than half of what the Pro Plan cost is and if you're just one operator I mean wait right like it's not very good yet and it's going to have a high failure rate for a while so I would wait a little bit of time until that's better overall deep seek chat gbt Pro not needed at the moment I would try out

the free plan and try out o03 mini and if you love it and want a hire plan go for Plus or team with that being said I hope you enjoyed this video and learned a little bit about what this new model can do for you thank you so much for watching and we'll see you in the next episode of the AI agents podcast bye