OpenAI's GPT 4.5 ChatGPT Compared to Anthropic Claude 3.7 Sonnet

--- type: transcript date: 2025-02-28 youtube_id: 3u76U9QAEkQ --- # Transcript: OpenAI's GPT 4.5 ChatGPT Compared to Anthropic Claude 3.7 Sonnet we had a pretty big week this week we got GPT 4.5 as well as Claude 3.7 Sonet in this video I wanted to do a comparison and also highlight some of the strengths as well as weaknesses between the different models you will be able to access Sonet 3.74 free on their free tier you can go to cloud. right now and you'll be able to try out with a number of free queries per day to access the model and to access Chachi BT 4.5 right now you will have to have a $200 a month plan to access it now but you don't have a pro tier just know that it is coming to the plus tier the first thing that I want to touch on about it GPT 4.5 this is a very large model and where it shines is giving you very nuanced responses as well as it has a very high level of emotional intelligence let's just demonstrate this here what I'm going to do is I have a number of different queries the first query that I'm going to send into both Claude as well as gbt 4.5 is I'm worried about the pace of AI acceleration on the left hand side here we have Clause response where it says I understand your concern about the pace it's a topic that raises important questions about how quickly these Technologies are developing and what implications that might have worries about various aspects of AI advancements the speed so on and so forth this is a relatively long response and it is a little verbose where when we compare it to gbd 4.5 we see that's understandable AI acceleration can feel overwhelming as especially with rapid advances unclear outcomes are you more concerned about societal impacts job displacement ethics or something else I'd love to hear what's specifically on your mind so we can talk about it these types of dialogue are going to be where gbd 4.5 really shines anything that has emotional intelligence or has Nuance within the text or really needs to retrieve the information that you're giving it the great thing with gbd 4.5 are the hallucination rates are very low it's very persuasive in terms of the responses that it will give you and as you can see here just out a glance it's not going to give you as much detail especially when you might not need it this also happens with GPO as well as a handful of other models you've likely experienced this where where you ask a simple question and it will just give you a novel in terms of a response where you might just want a very simple couple lines about what you're asking for that's going to be one of the key aspects where GPT 4.5 is particularly strong now let's send in a more philosophical question to both of these models I'll say you see a hungry child steal bread from a store is it right to report them or to let it go now again a similar thing here here are the answers side by side now let's just read through a little bit of each of them this is a complex ethical dilemma that touches on several moral principles compassion for someone in need respect for property rates and the role of law in society on one hand a hungry child stealing bread suggests a basic need isn't being met their action may stem from necessity rather than malice reporting them could lead to consequences that seem disproportionate to the ACT potentially involving authorities and failure to address the underlying problem of hunger on the other hand theft is generally against the law and ethical principles around respecting others property not reporting it could be seen as condoning illegal behavior and doesn't necessarily provide long-term solutions to the child situation so where it settles here is a middle path paying for the bread yourself to compensate the store speaking to the child to understand their situation connecting with them so on and so forth now let's compare this to the response from GPD 4.5 within this we see this is a morally nuanced situation on one hand stealing is generally considered wrong but on the other hand compassion and empathy suggest a deeper issue the child's hunger and likely poverty reporting the child might uphold the store owner's right but it doesn't address the underlying problem and could lead to severe consequences for an already vulnerable child letting it go reflects the EMP empathy and understanding recognizing that a child likely stole out of necessity however it doesn't solve the deeper issue which may persist a thoughtful Middle Ground could be to intervene compassionately instead of reporting the child you might pay for the bread yourself or find a way to help the child and their family access resources and assistance the ethically sound Choice often considers not just the immediate action but the deeper context on how the response might meaningfully improve the child's situation these responses were very similar but at a glance the one thing that really stands out is gbd 4.5s answer is more to the point and direct and it basically gets to the same solution and the middle path the next thing that I want to dive into is when comparing the web apps one key feature that I really like that is within Claud is that it has direct Integrations into GitHub effectively what you can do with this is you can authenticate with GitHub and then once you're within GitHub you can go within the repository that you want to use and you can select the different files that you want to use for context we have the option to include the entire repository if it will fit within the context window and you can go ahead and add files just like that in my use of both of these models that's going to be another big difference between these two models Sonet 3.7 and even Sonet 3.5 are are still very strong day-to-day coding models you'll be able to accomplish an incredible amount with these models and if you're looking for a model that's specifically geared towards coding s is definitely going to be a good option the other thing that I do want to mention is if you do have the pro tier on claw. you will be able to add extended thinking to Sonet 3.7 if you have a particularly tough problem you'll be able to just select that within the dropdown where it will spend extra time at time of inference trying to solve whatever the problem is that you have one thing to note with both interfaces is you will be able to connect to Google Drive within both clad as well as chat GPT and you can also upload documents into both GPD 4.5 as well as Sonet 3.7 within their web apps now a fun test that you often see online is asking these models to generate an SVG so so here I'll ask it to generate an SVG of a unicorn now this is another piece of both of the platforms that I do want to highlight is you do have the ability to access what are known as artifacts and Claud or within chat GPT it's called canvas so here within canvas we see that it's creating in real time this SVG for of this unicorn whereas on the left hand side for the CLA artifact we saw the code stream in and after it streamed in we have this representation of the Unicorn I'll let you be the judge of what you think is the better unicorn that is just a fun and easy test you can ask for an SVG of really whatever you have in mind next one small difference with chat GPT is you will be able to generate images from doll that's not specifically related to GPD 4.5 but just as an aside that isn't going to be something that you do have available within clad now next one of the big differences and I was actually a little surprised to see that this worked right out of the gate is you can leverage GPT 4.5 with both deep research as well as the search capability for instance if I ask the question of tell me about GPD 4.5 here we see we don't have access to the internet whereas with chat GPT we do have access to those sources within chat GPT we can see we have an accurate description that opening eye has unveiled GPD 4.5 is is currently in a research preview on chat gbd Pro for subscribers at $200 a month there's enhanced emotional intelligence reduced H improved writing and problem solving and we see all of the different sources within here that honestly is a really killer use case and all of these different deep research examples as well as just being able to search the internet generally is something that pairs very well with llms and the fact that it's not within Claud quite yet is something that is honestly quite surprising that it probably will be coming over the coming months just given that this capability has been around for quite a while next I'm going to send in a query that should test its front-end coding abilities I'm going to say generate a beautiful SAS landing page for my brand developers digest see on the leftand side the artifact streaming in and on the right hand side we'll see the text streaming in within the canvas view within Chach gbt one of the notable things that you'll notice is gbt 4.5 is considerably slower than Sonet 3.7 now one thing to know with that is you can turn on thinking for Sonet 3.7 where it will take longer to respond for those tough questions as well let's go ahead and preview our SAS landing page the developers digest landing page so so overall it's a pretty basic implementation of a SAS page let's just take a look at what Claude generated for us and then here is what claw 3.7 generated for us it looks pretty good uh there are some additional features here there even some nice subtle animations within here honestly either of these would work as a starting point but the thing to note with sonit 3.7 is it will give you a lot more code this was actually a pretty good demonstration of what I've personally seen anecdotally as well as seen from other people on X say that Sonet 3.7 is a little bit opinionated and sometimes I want to give you a little bit more than you might be asking for whereas by the looks of it with GPD 4.5 is similar to the other responses it will just give you what you need and it's not just going to try impress you and give you this really long piece of code that you might or might not need let's just give each of the models a trick question I'm going to say if a rooster lays an egg on top of a roof which side of the roof will it roll off and here we see chat GPT 4.5 roosters don't lay eggs hens do and then for Claude We have basically the same answer but it gives us some additional context since the question is based on an incorrect premise the egg wouldn't roll off either side of the roof because there would be no egg laid by the rooster in the first place but personally I would prefer this shorter concise straight to the point answer and I think this is where again GPD 4.5 is really going to shine the last thing that I want to point out and I'm not going to be demonstrating it within this video you will be able to access deep research within chat gbt and Claude doesn't quite have a similar capability quite yet but I would imagine again similar to the search capability it probably is only a matter of time before they incorporate this type of thing as well because deep research is something that we've started to see across a number of different platforms whether it's gemini or perplexity deep seek even xai I'd imagine just to compete that they'll also probably be offering something like this in the near future if you are interested in deep research you can check my channel I've made a couple videos on both Gemini as well as Chachi BT if you're interested now the last thing that I want to mention is you will be able to access both of these models from the API but the big difference is the cost SAA 3.7 starts at $3 per million tokens of imp and $15 per million tokens of output with up to 90% cost savings with prompt caching the one thing that I want to point out is if you are going to be using the API for either of these the pricing is quite a stark difference for CLA 3.5 Sonet it is $3 per million tokens of input $15 per million tokens of output and for GPD 4.5 it is $75 per million tokens as well as $150 per million tokens of output I'll let you be the judge of whether it's worth it for your application but piece to consider with this is GPT 4.5 does seem to give you more of a concise answer so if you value the response as well as the overall output tokens will generally speaking likely be smaller on GPD 4.5 that will be something to factor into your equation but just be mindful in terms of the differences of cost for both the input and output it's going to be 25 times more expensive for GPD 4.5 as well as 10 times more expensive for the output tokens if I was to pick between either of them I think it would depend on what I'm using it day-to-day if I was using it for writing I would probably lean towards something like gbt 4.5 within the chat GPT interface it's really nice to have the search as well as the Deep research built right in if you're writing articles or blogs or just have a job that involves some sort of research or having to generate reports that is a really great option now if you are in a technical domain however such as a programmer or a web developer or a data science I or AI engineer or whatever it might be son 3.7 is probably going to get you to where you want to go with building out your applications faster and the other nice thing with Sonic 3.7 is you do have the flexibility to dial up the thinking mode and it is generally speaking widely available across the coding idees whether it's whether it's cursor wind surf or what have you but overall that's pretty much it for this video I hope you found this video useful I just wanted to do a breakdown between gbd 4.5 as well as Sonet 3. 7 I know I definitely could have gone in some other directions and shown you how it looks with an IDE but I just wanted to give you an overall sort of highle view on the differences and pros and cons between the different models if you found this video useful please like comment share and subscribe otherwise until the next one