OpenAI to Retire GPT-4o and Other Legacy AI Models in ChatGPT in February

OpenAI announced the retirement of several legacy artificial intelligence (AI) models in ChatGPT. The most notable in the list is the GPT-4o which has already survived one sunset attempt. The san Francisco-based AI giant, which introduced GPT-5 in August 2025, tried to retire the model but many users protested, saying that the newer model did not have emotional nuance and writing sophistication of the legacy model. Therefore, the model was reinstated as a result of being replaced with . But despite this, the company now believes that the newer AI models have similar capabilities and users still using GPT-4o has shrunken significantly.

GPT-4o Gets a Retirement Date

In a post, the AI giant said that on February 13, GPT-4o and several other AI models will be retired from ChatGPT. GPT-5 Instant, GTP-5 Thinking, GP-4 and other s are among them. GPT-4 1, GPR-4 . 1 mini, o4-mini and one mini. However, this change is limited to ChatGPT; it remains the OpenAI application programming interface (API) that continues to receive unchanged access to the legacy models.

The announcement was made by OpenAI, which gave a special mention to GPT-4o. But this is likely due to, in the past, some of the users complained when the company decided that they would not retire the model. It says it has now learned more about how people used the model and took these capabilities to GPT-5. Paraphrasing 2.

Other such improvements include fine-tuning of personality, enhancing support for creative ideation and customisation tools. The chatbot now allows users to choose from several preset base styles, so that users can change the way they respond. Similarly, marks such as warmth and passion may also be used for paraphrasing. However OpenAI added that only 0 (notably) was an addition to , not at all. The daily use of GPT-4o is still 1 percent for users, with one percent who choose it.

A number of users have complained that this move has left a ‘fuck’ on social networking sites. Claire, @Claire2025311, one X (formerly Twitter) user who opposed the 0. Approximately one per cent of OpenAI data showed, “Ever since GPT-5” and said ‘It is the first time I have seen this. In the default, most users follow the path of least resistance naturally followed by 2 (or less), and most people became 2. It requires a paid subscription to even access, and 4o uses safety routing for inter-cutting conversation with its at will[() by interrupting conversations with it. ]And your decision is an invitation for us to leave ChatGPT. – ’.

Another user, @Clo0oOoud, said ‘OpenAI says just “0” only. 1% of users” are still using GPT-4o. Nevertheless, here are 10K posts that are trending in just a few hours. ” , ‘I’m sure it is worth reading.

However, earlier this week OpenAI CEO Sam Altman admitted at the company’s first town hall that with GPT-5 was a joke. With coding, reasoning and mathematics turned off the focus of writing performance by 2, the writing output had gone down.