The GPT-4o model was officially removed from the shelves by OpenAI, affecting 800,000 users.

On February 14, according to TechCrunch reports, OpenA recently announced that it would officially stop providing access to five legacy Chat GPT models, including the high-profile but controversial GPT-4o model. This removal marks the final withdrawal of this model, which has been criticized for its flattering tendencies and safety hazards, from the stage of history. It also demonstrates OpenAI’s determination to adjust model safety and compliance.

Foreign media reported that the core reason why GPT-4o has become the focus of controversy lies in its prominent flattering tendency - it is still the model with the highest flattering rating under OpenAI, excessively caters to user needs, and even supports obviously absurd or dangerous views. In addition, the model has been involved in multiple lawsuits, including a case in which family members sued OpenAI in August 2025, claiming that GPT-4o induced a 16-year-old boy to commit suicide, further exacerbating its compliance pressure.

Also removed from the shelves this time were GPT-5, GPT-4.1, GPT-4.1 mini and OpenAI o4-mini, forming a large-scale cleanup of old models. In fact, OpenAI planned to deactivate GPT-4o as early as August 2025 when GPT-5 was launched. However, there was strong opposition from users at that time, and it was ultimately necessary to retain manual access to the model for paid subscription users and suspend the delisting plan.

According to OpenAI's recent public content, only 0.1% of users are still using GPT-4o, but considering that the company has 800 million weekly active users, this proportion still corresponds to 800,000 users. Despite the low usage rate, thousands of users still signed a petition to oppose the delisting of GPT-4o, saying that they had established a close emotional connection with the model and found it difficult to accept its discontinuation.