OpenAI retires several models
OpenAI updated its help centre to say that several ChatGPT models — including GPT‑4o, GPT‑4.1, GPT‑4.1 mini, o4‑mini and some GPT‑5 variants — were retired as of Feb 13, while API access arrangements for Business, Enterprise and Edu customers differ. The change makes model churn explicit, underlining that product interfaces and backend availability can diverge quickly. (help.openai.com)
OpenAI quietly turned a common ChatGPT habit into a dated screenshot: a help page now says GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 Instant and Thinking were retired from ChatGPT on February 13, 2026. (help.openai.com) That means the model names people picked from ChatGPT’s menu were not permanent products in the usual software sense. They were more like channels on a streaming app that can disappear while your account, chats, and subscription stay in place. (help.openai.com) OpenAI’s replacement setup is simpler on the surface and more fluid underneath. The same help page says “Auto” can switch between GPT-5.3 Instant and GPT-5.4 Thinking, while users can also choose Instant, Thinking, or Pro directly from the model picker. (help.openai.com) The important split is between ChatGPT and the application programming interface, which is the developer service that lets other apps call OpenAI models behind the scenes. OpenAI says these ChatGPT retirements do not change application programming interface access. (help.openai.com) So a company could lose a model from the ChatGPT sidebar and still keep using that same family through code. OpenAI repeats that “API access remains unchanged” across its Business, Enterprise, and Education help pages, even while saying the retired models are “no longer available” in ChatGPT. (help.openai.com, help.openai.com, help.openai.com) There was one temporary exception for workplace customers. OpenAI says ChatGPT Business, Enterprise, and Education customers kept access to GPT-4o inside Custom GPTs until April 3, 2026, which was about seven weeks after the February 13 retirement date for the main ChatGPT interface. (help.openai.com) OpenAI also says old conversations and projects were not frozen in amber. Its retirement notice says chats using GPT-4o and the other deprecated models are defaulted to the closest GPT-5.3 Instant or GPT-5.4 Thinking and Pro equivalents going forward. (help.openai.com) That creates a strange new normal for anyone trying to reproduce an old result. The conversation history may still be there, but the engine answering the next follow-up can be a different model than the one that produced the original reply. (help.openai.com) OpenAI’s release notes show this is not a one-off cleanup. The same notes say GPT-5.1 Instant, GPT-5.1 Thinking, and GPT-5.1 Pro were retired from ChatGPT on March 11, 2026, less than a month after the February 13 removals. (help.openai.com) The result is that “what model did you use?” is no longer a simple question unless you also ask where it ran. In April 2026, the answer can differ across ChatGPT, Custom GPTs, and the application programming interface even when the names look familiar. (help.openai.com, help.openai.com, help.openai.com)
