4o has a very strong artificial vibe. It feels a bit "autistic" (probably a bad analogy but couldn't find a better word to describe what I mean): you feel bad ? must say sorry then give a TODO list on how to feel better.
4.5 still feels a bit artificial but somehow also more emotionally connected. It removed the weird "bullet point lists of things to do" and focused on the emotional part; which is also longer than 4o
If I am talking to a human I would definitely expect him/her to react more like 4.5 than like 4o. If the first sentence that comes out of their mouth after I explain them that I feel bad is "here is a list of things you might consider", I will find it strange. We can reach that point but it's usually after a bit more talk; human kinda need that process, and it feels like 4.5 understands that better than 4o.
Now of course which one is "better" really depends on the context; what you expect of the model and how you intend to use is.
Until now every single OpenAI update on the main series has always been a strict improvement over the previous model. Cost aside, there wasn't really any reason to keep using 3.5 when 4 got released. This is not the case here; even assuming unlimited money you still might wanna select 4o in the dropdown sometimes instead of 4.5.
4.5 still feels a bit artificial but somehow also more emotionally connected. It removed the weird "bullet point lists of things to do" and focused on the emotional part; which is also longer than 4o
If I am talking to a human I would definitely expect him/her to react more like 4.5 than like 4o. If the first sentence that comes out of their mouth after I explain them that I feel bad is "here is a list of things you might consider", I will find it strange. We can reach that point but it's usually after a bit more talk; human kinda need that process, and it feels like 4.5 understands that better than 4o.
Now of course which one is "better" really depends on the context; what you expect of the model and how you intend to use is. Until now every single OpenAI update on the main series has always been a strict improvement over the previous model. Cost aside, there wasn't really any reason to keep using 3.5 when 4 got released. This is not the case here; even assuming unlimited money you still might wanna select 4o in the dropdown sometimes instead of 4.5.