OpenAI says retiring older models helps it improve newer models without changing its API.
Several ChatGPT AI models, including the popular GPT-4o, will be retired next month as OpenAI focuses on newer versions of GPT-5, which most users currently use.
The organization stated that it will discontinue GPT-4o from ChatGPT on February 13, along with GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini. This is because the company realized that user behavior has changed, with most people using GPT-5.2 for their daily needs.
OpenAI launched GPT-4o in May 2024. The model was designed to convey a more natural, human feel in chat sessions. It had a warmer tone compared to other ChatGPT models and attracted people seeking a more expressive, creative experience.
This strategy helped GPT-4o attract a loyal group of Plus and Pro users. These users used the model for ideas, writing, and brainstorming, and reported that it was easier to talk to and better for open-ended work than what was available at the time.
However, over time, users’ interactions with ChatGPT have evolved. Currently, according to OpenAI, “about 0.1% of users choose GPT-4o daily,” with most users now using “GPT-5.2, which has become the default model for most daily tasks.”
Considering that very few people are using GPT-4o these days, OpenAI claims that supporting this model distracts from efforts to improve the models most people actually use. The time and resources used to support these older models can instead be devoted to improving the newer models, making them faster, more reliable, and easier to use.
By discontinuing GPT-4o and other less frequently used models, OpenAI claims it can devote more effort to improving the performance, reliability, and creativity of its new models, especially amid increasing demand for better AI tools.
The company further noted that GPT-5.2 delivers more powerful creative output and greater control over the model’s personality. Users can also fine-tune the GPT-5.2 model’s response. As a result, the need to maintain the older versions of the model, such as the GPT-4o, is reduced.
GPT-4o is now ready for retirement after a period characterized by talks that began in August. This followed OpenAI’s decision to withdraw the model shortly after releasing the latest version, dubbed GPT-5. The decision was immediately opposed by the users, who felt it had come too early.
A lot of people back then felt that GPT-5 was a regression because it didn’t have the warmth, grasp, and flowing conversation that they were used to in GPT-4o.
OpenAI decided to reverse the decision and restore GPT-4o for paying subscribers. They stated that many of their subscribers required additional time to transition their critical tasks to the new models without any disruption.
However, following this reversal, OpenAI’s Chief Executive Officer, Sam Altman, promised users that the organization would give them “plenty of notice” before making a permanent decision regarding the closure of GPT-4o.
Despite GPT-4o’s presence in the market, OpenAI has been paying close attention to the feedback it has been receiving and using it to make subsequent versions even better.
The focus has been on refining GPT-5.1 and GPT-5.2 in terms of tone, creativity, and conversational style. The company states that the new versions allow users to adjust ChatGPT’s personality, letting people control the level of warmth or enthusiasm in the chat rather than being stuck with a single default setting.
OpenAI also said the previous debate on the older model touched on broader issues, such as how people tend to get emotionally attached to AI and the possibility that AI might give overly agreeable responses. They emphasized that they are working on solving these issues with the new models and versions of ChatGPT.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It’s free.
