OpenAI fragments model access
OpenAI says several older models have been retired from ChatGPT while GPT‑4o remains available to ChatGPT Business, Enterprise and Edu customers through Custom GPTs and projects, with API access unchanged. The move underscores increasing segmentation of model availability by channel and customer tier. (help.openai.com)
OpenAI has split model access more sharply inside ChatGPT: several older models are gone from regular chats, while some business customers kept a narrower path to GPT-4o for a few extra weeks. (help.openai.com) OpenAI’s help center says GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 Instant and GPT-5 Thinking were retired from ChatGPT on February 13, 2026. A separate note says GPT-5.1 models were retired on March 11, 2026 across normal chats and GPTs. (help.openai.com, help.openai.com) The carveout was limited. OpenAI said ChatGPT Business, Enterprise, and Edu customers could still use GPT-4o inside Custom GPTs until April 3, 2026, even after it disappeared from standard ChatGPT use. (help.openai.com, help.openai.com) That exception has now ended. OpenAI’s retirement notice says GPT-4o is “fully retired across all plans” after April 3, 2026. (help.openai.com) The split matters because “ChatGPT” is no longer one uniform product. OpenAI’s own documentation now distinguishes between normal chats, Custom GPTs, projects, workspace controls, and the application programming interface, with different model rules in each place. (help.openai.com, help.openai.com) OpenAI also said the retirements did not change developer access. Its help pages state that the affected models remain available through the OpenAI application programming interface, and that the company will give advance notice before future application programming interface retirements. (help.openai.com) For workplace customers, OpenAI is steering users to newer defaults. The “Legacy Model Access” page says Enterprise and Edu workspaces are automatically using GPT-5.2 for ChatGPT, while deprecated Custom GPTs move to the “closest-matching currently available model.” (help.openai.com, help.openai.com) OpenAI has been making the same distinction in other product pages. The Business rate card, model-limits pages, and GPT creation docs all repeat the same retirement language, tying model availability to plan type and feature surface rather than to a single company-wide list. (help.openai.com, help.openai.com, help.openai.com) The practical result is simpler than the documentation: the model you can use now depends on where you use OpenAI’s tools, what plan you pay for, and whether you are in ChatGPT or the application programming interface. (help.openai.com, help.openai.com)
