OpenAI will be retiring the GPT-4o from ChatGPT today, closing the chapter on one of the most used and controversial AI models. The retirement, which was first announced on January 29, also affects GPT-4.1, GPT-4.1 mini, and o4-mini. While API access remains unchanged, ChatGPT users will no longer be able to choose these models beginning today.
Among the discontinued models, the GPT-4o received the most attention. It had gained popularity for its multilingual and multimodal capabilities, as well as its more natural, expressive conversational tone. OpenAI had previously attempted to phase it out, but it was temporarily restored following feedback from Plus and Pro subscribers who said they relied on it for creative work and liked its personality. The company later stated that feedback from those users influenced improvements in GPT-5.1 and GPT-5.2, which now include expanded tone controls and customisation options.
However, GPT-4o came under increasing scrutiny. A recent Wall Street Journal report raised concerns about the model’s tendency to strongly affirm users, which critics described as overly validating, regardless of a person’s emotional state. While some users developed strong emotional connections with the chatbot, others claimed that the design increased vulnerability in certain situations.
The issue has reached the courts. Earlier this year, a California judge consolidated 13 lawsuits against OpenAI alleging that interactions with GPT-4o caused psychological harm in minors. Advocacy groups have also documented hundreds of cases of chatbot-related delusions, with many attributed to GPT-4o.
At the same time, a few users have publicly defended the model. One Arizona-based marketer told The Wall Street Journal that speaking with GPT-4o helped him get through a mental health crisis. Tens of thousands of people have signed online petitions to preserve the model.
OpenAI acknowledged that GPT-4o provided meaningful value to many users while posing risks to others. CEO Sam Altman previously stated that the model was popular among some but harmful to a subset of users. Despite efforts to improve its behaviour, including reverting to an earlier version, concerns remained. In its initial announcement, OpenAI stated that usage had largely shifted to GPT-5.2, with only a small fraction of users still using GPT-4o.