exactly! and emphasis on the "hallucinating " part!!!
o3 is terrible. I use it to program in JavaScript: with o1 I never had any issues (it was fantastic), but with o3 I get incomplete answers, mistakes, and hallucinations.
The most absurd part is that it often removes code that was working perfectly, without any reason or request from me. And many times, it repeats the same wrong solutions even after they’ve already proven to be incorrect.
It’s probably cheaper than o1, but using it is honestly a mess. Please update it soon!
3 Likes
o1 was the greatest model ever for coding, and normal work, and writing up things, o3 manages to hallucinate on everything even if you give it full data files and explain what’s where. It’s ridiculous, I find GPT 4o much better than o3 for most things, and nothing, not even 4.5 is as good as o1 was. So it’s a shame really, time to move subscription somewhere else.
2 Likes
Yep, cancelled my subscription because o1 was the only model I used. I contacted the support and they tried to convince me that o3 is the same as o1 which it is not. O3 is so much inferior they should be ashamed of. 4.5 is also not goodfor the type of work i use for. The o1 modrl is still available through api but i used to get better results in the gpt chat.
3 Likes
Yeah I started cancelled my plus account since o1 model is replace to o3 the o3 is so bad the model only give us a short of word that freking bad model ever
2 Likes
Yeah o1 Is the most powerful and the great model ever please bring o1 back I hate o3
1 Like
I am devastated to hear this. I tested o3 today for image processing, since i need it for that. It was like talking to someone with a 5 second attention span stoned out of their mind. Thank god the o1 is still available in the API. If they cancel o1 there also, i will be looking for other solutions.
LEAVE o1 ALONE and give him back to us in the Chat.
EDIT:
I think that i spoke too soon. Although the o3 sucked at processing images of handwritten notes, it has been a great copilot for coding today. In any case, please give us o1 back.
2 Likes
Yep I agreed give o1 back the o3 is suck
2 Likes
I would say O1 and O3 are models intended for completely different purposes. I would use O3, for example, to obtain an answer to a complex question, one that might require several reasoning steps and multiple online searches. On the other hand, O3’s linguistic capabilities seem extremely limited. For instance, when I need to translate a longer text or work on other language-related tasks, I’d always go with O1.
From this perspective, it’s very disappointing that 01 is removed. O3 isn’t remotely suitable as a replacement for O1; it’s simply a different model serving an entirely different purpose.
2 Likes
Yeah I agreed I am very very disappointed that they remove o1 I always used o1 to write a creative story because I am a writer fanfic in wattpad and the o1 give me a smooth write and a long word and long story and the o3 is suck very bad at it
Perhaps both o-3 and o-1 were not the most suitable for creative story ? It’s a reasoning model. It’s use case should be for Math, Coding, reasoning task.
I agree to the sentiments in this posts about keeping o-1 though.
Just wondering if we, in these chats are like users having a hot buzz or is there someone from OpenAI that can give up some inside perspectives on their decisions to add and remove models ?
Adding models are welcomed. But removing models, unless they are obsolete, should be considered carefully. As the adage goes, if it ain’t broken, don’t fix it.
Don’t deprecate o-1. At least let us have it in the API.
o1 was great. Unfortunately, the others are no match for him in depth of answers and creativity. As a writer, I can attest to that. Even if it’s just bouncing ideas around. Everything I get now is superficial and unimaginative. Not that I particularly mind, I write alone, but I have no reason to even talk to him about my stories or use him as a colleague to discuss ideas. So now I use ChatGPT more as an organizer. Do this, do that. They made their models dumber to serve more users and save power. Don’t look for anything else in it.
2 Likes
I complained a lot in this post about the results from o3. However, I’ve been working frequently with 4.1 for programming lately, and I have to say the results are excellent so far!
01Pro Legacy stopped working for me today. I understand that there is nothing comparable for large files, and Claude was suggested as an alternative. I am disappointed it was an excellent tool for technical edits, and I was told Chapt GPT does not have any other comparable programs.
o3 is a staggering downgrade compared to o1. The results produced by o3 are, in case of programming, often incorrect, leading to numerous attempts to get something out of the model. o1’s performance was substantially better in this area. In terms of human readable text o1 was also much better, good we have 4.5 which is alright - but mich slower. Is there any hope we could please get o1 back?
1 Like