Pour out a cold one for ChatGPT-4o. Parent company OpenAI announced on Thursday it is retiring several of its older models, including GPT-5, GPT-4o, GPT-4.1, GPT-4.1 mini and o4-mini. The last day to use these models is Friday, Feb. 13.
Normally, deprecation of older AI models wouldn’t warrant scrutiny — or even a news story — but ChatGPT-4o is special. It’s weird to say, but for many ChatGPT fans, that model is a favorite.
When the company released GPT-5 last year, it removed GPT-4o from the available options. Many were vexed by how short and unfriendly the new model was compared to GPT-4o. Some were livid, frustrated that their go-to model was disappearing overnight, with little they could do to get it back. OpenAI brought it back later that week.
Some experts were worried that GPT-4o and other models’ friendliness crossed the line into sycophancy. AI sycophancy occurs when models are overly affectionate, becoming digital yes-men that potentially validate users’ dangerous ideas.
That’s likely why OpenAI published a lengthy blog post this week about its thinking behind removing older models like GPT-4o.
«We know that losing access to GPT‑4o will feel frustrating for some users, and we didn’t make this decision lightly,» the company wrote in a blog. «Retiring models is never easy, but it allows us to focus on improving the models most people use today.»
OpenAI said that only 0.1% of its users regularly use GPT-4o to run tasks. That amounts to about 800,000 users, based on OpenAI’s most recent count of 800 million weekly active users in its 2025 enterprise report.
OpenAI is betting that enough time has passed to avoid angering GPT-4o loyalists as it did before. We’ll soon see if the new GPT-5 models, GPT-5.1 and GPT-5.2, have won over enough fans.
