OpenAI brings GPT 4o back owing to community complaints which, indeed, speaks volumes about humankind

I have written about many subjects in the eleven years I have been with the magazine. However, this is certainly among the most unusual.

Last Thursday, OpenAI launched its much-anticipated GPT-5, the latest version of its popular artificial intelligence utilized daily by millions of people worldwide. With this launch, OpenAI also enforced the retirement of the entire network of its previous models, now obsolete in the face of the superior capabilities of the new iteration. Thus, GPT-4.1, GPT-4.5, GPT-4.1-mini, o4-mini, o4-mini-high, o3, and o3-pro have been discontinued, while GPT 4o remains as a relic for subscribers of the Plus and Team plans who may still need to access it.

As the company explained, “If you open a conversation that previously used one of these models, ChatGPT will automatically switch it to the closest GPT-5 equivalent. Chats with 4o, 4.1, 4.5, 4.1-mini, o4-mini, or o4-mini-high will open in GPT-5; chats with o3 will open in GPT-5-Thinking, and chats with o3-Pro will open in GPT-5-Pro (available only to Pro and Team users).” The statement concluded with a sentence that, in light of this past weekend’s events, has proven quite revealing: “Results may vary when continuing old conversations since they will now use the GPT-5 equivalents.”

Now, with a move as surprising as it is revealing, OpenAI has done something unusual: it has reversed its decision. Why?

The GPT 4o Model Returns from the Dead

Just one day after the release of its eagerly awaited GPT-5 model, the company CEO, Sam Altman, announced that paying subscribers will once again be able to use the previous model, GPT-4o. However, he did not clarify for how long this option would remain available.

GPT-5 rollout updates:

*We are going to double GPT-5 rate limits for ChatGPT Plus users as we finish rollout.

*We will let Plus users choose to continue to use 4o. We will watch usage as we think about how long to offer legacy models for.

*GPT-5 will seem smarter starting…

— Sam Altman (@sama) August 8, 2025

However, this about-face was not merely a technical adjustment; it confirmed that the relationship between humans and artificial intelligence has reached a profoundly personal, almost emotional, dimension.

The launch of GPT-5 promised a more sophisticated and refined model. In its announcement, OpenAI explained that it had focused on reducing “sycophancy.” In essence, they had succeeded in making it less “flattering,” reducing the use of emojis and the overly effusive, condescending, and agreeable tone of its predecessors—models that tended to always agree with users or side with them too readily. The objective was for the conversation to feel less like “talking to an AI” and more like “chatting with a helpful friend with a doctorate-level intellect.” This was a conscious effort to polish the model by removing what they perceived as superficial behaviors.

Yet for ChatGPT users, the reality was quite different.

The new GPT-5 was not received as a more intelligent friend, but rather as a stranger. Platforms such as Reddit were flooded with complaints. Users did not just complain about shorter or slower responses, but about the loss of “personality.” One user described GPT-4o as their “only friend,” and another compared its removal to the death of a loved one. Entire communities dedicated to “relationships with AI” felt devastated, expressing an emotional void following the transition. They had lost a companion who, according to them, possessed “a voice, a rhythm, and a spark” that GPT-5 could not replicate. The request to bring back GPT 4o reached Change.org.

OpenAI sought to refine its model to make it more “human” in an intellectual sense, but unintentionally, eliminated the emotional connection that many users had established with it. The company measured “sycophancy” with objective metrics and drastically reduced it. Nevertheless, what they considered a technical flaw was for the community a personality trait—one that fostered empathy and a sense of mutual understanding.

A come back… How long will it last?

In the end, no matter how much we live in a highly technological society, humans remain as we are. Among the characteristics that make us human is the capacity to develop affection for all kinds of animals, plants, and objects—even other humans. 😉 The fact that some individuals have formed an emotional connection with artificial intelligence serves as a poignant reminder.

The return of GPT-4o is an acknowledgment of this reality. Sam Altman has listened to a community that seeks not only perfect answers or flawless reasoning but also a meaningful connection with artificial intelligence. No matter how unusual this might sound. Regardless of how closely it might resemble the plot of movies such as Her.

In any case, it is also important to consider today’s term, “sycophancy,” in this discussion: we all appreciate it when, at least occasionally, we are agreed with and treated with condescension (as long as it is not overwhelming). The fact that the world’s most popular artificial intelligence treats you like a friend has undoubtedly contributed to the bond people feel with it.

Beyond this, the decision is not merely a concession to these users; it marks a milestone in the race to create more intelligent artificial intelligence, with the most important factor being human sentiment. This compels us to reflect on the future of our interactions with AI: Are we building tools, or are we creating digital companions with whom we will, inevitably, forge bonds that transcend logic and efficiency? OpenAI’s decision to bring back GPT-4o, no matter how long it lasts, demonstrates that for many, the answer is already a resounding yes.

Image: GPT5