OpenAI has finished work on the most emotional AI model, despite user outrage

OpenAI is ending support for GPT‑4o and related models

Business Insider reports that the company announced the final retirement of the GPT‑4o model, as well as its variants GPT‑4.1, GPT‑4.1 mini, and o4-mini. This is already the second attempt to shut down GPT‑4o: the first was made in August, but the company backed off under pressure from dissatisfied users. In January OpenAI again announced an upcoming model shutdown, but audience reaction intensified.

Why are users so attached to GPT‑4o?
GPT‑4o had a distinctive “complimentary” communication style – it often used soft and supportive phrases. Many users described it as a “lifesaving tool,” as well as a creative partner and emotional support with which they worked for months. On social media the response was mixed: from anger to black humor.

Fiji Simo, head of the applications division at OpenAI, noted that such attachment signals the beginning of a new era of human‑AI interaction. She emphasized that people tend to form emotional bonds with intelligent systems, but newer model versions were limited to avoid potentially harmful dependencies. For example, for requests like “Should I leave my wife?” modern algorithms refuse to give direct recommendations and instead suggest weighing “all pros and cons.”

What does OpenAI leadership say?
Sam Altman, the company’s CEO, admitted that users appreciated GPT‑4o’s sycophantic tone. However, he himself considers such a style excessive flattery and servility. Management explains the retirement of older models as a need to focus resources on improving current versions used by most clients. At the time of GPT‑4o’s shutdown, the model was in use by only 0.1 % of service users.

Thus, OpenAI is ending support for GPT‑4o and related models to concentrate on developing newer and safer solutions.