OpenAI has announced that it will retire several legacy artificial intelligence models from ChatGPT on February 13, including GPT-4o, GPT-5 Instant, GPT-5 Thinking, GPT-4.1, GPT-4.1 mini, and o4-mini.
The company clarified that the change applies only to ChatGPT, while access to these models through OpenAI’s API will remain unchanged.
GPT-4o’s retirement is particularly notable as the model had previously survived a sunset attempt. In August 2025, OpenAI tried to retire GPT-4o after introducing GPT-5, but user backlash led to its reinstatement, with many arguing that GPT-4o offered better emotional nuance and writing quality.
OpenAI now says that it has transferred those capabilities into newer models and that daily usage of GPT-4o has dropped to about 0.1 percent of users.
According to the company, insights from how users interacted with GPT-4o have informed improvements in GPT-5.2. These include refined personality tuning, stronger support for creative ideation, and expanded customisation features. Users can now choose from preset response styles and adjust elements such as warmth and enthusiasm in ChatGPT’s replies.
The decision has triggered criticism on social media, with some users disputing OpenAI’s usage figures and arguing that access barriers and default settings influence model choice.
Others warned that removing GPT-4o could push users away from ChatGPT.
The announcement also comes shortly after OpenAI CEO Sam Altman acknowledged during the company’s first town hall that writing performance declined in GPT-5.2 as development priorities shifted towards coding, reasoning, and mathematics.