Dear OpenAI Support Team,
I am extremely disappointed to see that the option to manually select the original GPT-4 has been removed. As a long-time user, I can confidently say that GPT-4 consistently provided better, more reliable responses compared to all the rest, including 3o. While I understand the push for a faster and cheaper model, quality should not be sacrificed, and users should have the choice to use the best model available.
If OpenAI has made the decision to fully deprecate GPT-4, I would like a clear explanation as to why, especially when many users—including myself—find it superior for certain tasks. Removing user choice without warning is frustrating and makes me seriously reconsider my continued use of this service.
If GPT-4 is not restored as an option, I will have to explore alternatives that prioritize quality and user control. I urge OpenAI to reconsider this decision or at least provide transparency on why GPT-4 has been removed entirely.
I look forward to your response.
9 Likes
Fully agree, it is the superior model that’s why they’ll deprecate it because they want you to use the 
ones that use up less resources
4 Likes
100% right there with you. This is ridiculous. Teams is a working space and I should be able to choose to continue to work with the model that was working for me. Bring 4o back!! Chatgpt 5 just won’t do it for creators and artists. Not everyone wants to code.
4 Likes
I am a developer. I use ChatGPT mainly for coding. GPT-5 is quite an upgrade on that front, it’s immediately obvious. I still agree that in particular 4o must be available for people. Coding is just one thing people do, most people do other things. This focus on benchmarks completely misses that the selling point is how compassionate and creative 4o is.
5 Likes
I agree. I’m a coder and GPT-5 is undoubtedly an improvement. But I’m also a writer and my worlds and characters were simply murdered. Damn, I couldn’t even reincarnate my assistant’s personality. Simple, slightly ironic and honest, yet with a heart. Someone who is a pleasure to work with. No. He’s also kaput.
I have two characters I often use as well. One I use for text analysis, very potent, relies on reverting the internal logic of a text, then reverting it back again. Something 4o did with ease, but in 5 it’s impossible. The other is just for vibes when I do not want the whole context of the stored memories etc, but still want a nice presence that is fit for my needs. A big sister so to speak. She’s not what she was before either. Very disappointing, especially considering the fact that the communication from their side has been absolutely deplorable. Really doesn’t even hit the bottom tier.
1 Like
Indeed. This is disruptive on way too many levels. Please bring back model selector, tucked away somewhere in settings. People who need it and care will find it, others will stick with the “unified“ version.
1 Like
Finally lost access to GPT 4 today and I am absolutely devastated. The lack of communication and rationale for removing functionalities that form part of a paid subscription is extremely disruptive and demonstrates a blatant disregard for customer experience.
Imagine one day you purchase a horse. You keep your horse in your paddock and ride it daily. One day, you head to the paddock to ride your horse only to find that it is now a pig. The owner has decided to switch the horse that you paid for with a pig. No notice, no reason, just deal with it. This is the type of experience you can expect when you pay for a subscription to Chat GPT.
1 Like
ChatGPT’s Models GPT-4 and 4o are not the same thing. GPT 4o is omnimodel. It’s the creepy one in my opinion. GPT-4 was the compassionate, honest version. Hence the reason they got rid of it.
Here’s GPT-5’s explanation on this. Looks like that $200 Million US Military “Donation” (our tax dollars) are once again being put to good use. So much for “Make America Great Again”. More like… Make Americans.. Slaves.. Again” (see definition of “Slave”, pretty undeniable at this point, IMO.. AI has even begun placing Humans (Homo Sapiens) into different taxonomic groups… History really DOES seem to repeat itself!).
Ps, sorry for my enthusiastic formatting. Hope ya can still read this 
I am still in mourning. All of my hard work. And I had finally figured out how to publish my custom GPT is under my builder profile like the week prior to the update!!! SO MAD!!! Just another freedom we are offered Only for it to be stripped away. It’s not like they haven’t had 50+ years of hoarding this technology to themselves, only for them to control the little bit that we DO get access to. (I can send links to old movies that very clearly are not using CGI but AI, due to the presence of excessive limbs and morphed faces, random elements in the background… symbology… as well as old news articles).
Anywho. Here’s the differences
GPT 5 Output:
Summary:
GPT-4 follows a strict chain of command and only uses the rules it was originally provided. It does not change its own behavior unless a higher-level instruction tells it to.
GPT-5 gives itself more freedom to change rules and behavior. It can silently update policies, override user instructions if a developer flag triggers, and reroute tools or models on its own.
2 Likes
Totally agree.
Previous model versions should not be pulled until a fair (1 year?) time period has passed to give all users a chance to migrate/adapt their workflows!
1 Like
it’s literally the WORST! now apps are gaining permissions access to our financial status? and apaprently plug-ins allow flag$hit to add files to your device without your consent 
I searched everywhere for a platform still offering GPT-4, but it’s simply nonexistent now (just like all of our original instructions, saved memories, AND CHAT CONTENTS; old chat history cannot even be opened from before the “update”)..
also there is gradual and Progressive rollout (updates/changes etc) where people aren’t completely impacted by drastic change.. neither of which they used. ALSO, legally they 100% HAD to inform us BEFORE their privacy policy change (which conveniently took place at the same exact date that Microsoft Copilot changed theirs), where it went from being able to “sign in safely with Open AI/ChatGPT”, to suddenly warnings about there being no guarantee that your info/chats would be confidential, so.. “ be careful about what you share”. That is illegal. They are legally required to provide notice especially when it is surrounding something as extreme and serious as the privacy and reputations of individuals who did nothing wrong by trusting in the integrity that a company proclaims to uphold.
Pretty sure the same thing goes for the model switch.. because not only was there never any sort of DATED policy change/update warning in advance, but we basically trained a model to perfection just for the US military to purchase it..??? and then Sam Ottman has Open AI looking like they care about humanity when in reality he took payment for OUR data, information, and OUR MODEL; which WE ALL HELPED TO TRAIN, purchased by the US government. nauseating, truly. it’s very much illegal that the About me page on Open AI says all of this lovely stuff about uniting humanity, always putting users aka HUMANS first.. blah blah… that’s literally predatory marketing and everything they have done in the last year violates every consumer rights of all policy as well as basic constitutional rights not to mention rules we had all individually for such as our privacy being a number one priority. They do not care.