OpenAI Model Lineup Change
OpenAI updated ChatGPT’s public model lineup and retired several named variants while keeping API access unchanged for Business, Enterprise and Education customers, highlighting that model availability is now part of commercial packaging. That shift means entitlements and migration clarity are becoming central to renewals and upsells for firms building on these platforms. (help.openai.com)
OpenAI quietly changed the part of ChatGPT most people actually touch: the model picker. In the new setup, GPT-5.3 became the default for logged-in users, GPT-5.3 Instant became the everyday fast option, and GPT-5.4 Thinking became the heavy-duty reasoning option. (help.openai.com) At the same time, OpenAI retired a long list of older ChatGPT names on February 13, 2026: GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 Instant and Thinking. Those names disappeared from ChatGPT itself, even though the underlying application programming interface, or API, stayed available. (help.openai.com) That split is the whole story. ChatGPT is the storefront people click on, while the application programming interface is the back-room connection companies wire into products, and OpenAI is now treating those two layers differently. (help.openai.com) If you are a normal ChatGPT user, the change looks simple: fewer names, more auto-switching, less manual model shopping. OpenAI’s help page says GPT-5.3 in ChatGPT is now “a single auto-switching system” that combines models behind one default experience. (help.openai.com) If you are a company paying for ChatGPT Business, Enterprise, or Education, the change looks less simple, because model access is now tied to plan rules, seat types, and transition windows. OpenAI’s Business and Enterprise help pages say API access remains unchanged, but ChatGPT access does not. (help.openai.com 1) (help.openai.com 2) OpenAI even left one temporary bridge in place: Business, Enterprise, and Education customers kept access to GPT-4o inside Custom GPTs until April 3, 2026, after the broader February 13 retirement. That is the kind of detail procurement teams look for, because a two-month grace period is very different from a same-day cutoff. (help.openai.com) The packaging is getting more explicit in other ways too. On April 2, 2026, OpenAI added two seat types for ChatGPT Business and ChatGPT Enterprise: a standard ChatGPT seat and a Codex-only seat. (help.openai.com) That means “which model do we get?” is no longer just a product question. It is becoming a contract question, because access can depend on whether a user has a standard seat, a Codex-only seat, legacy access turned on, or a workspace-level setting. (help.openai.com 1) (help.openai.com 2) OpenAI’s legacy access page makes that plain. It says Enterprise and Education workspaces can temporarily turn on older models such as GPT-4o, OpenAI o3, OpenAI o3 Pro, GPT-4.1, and GPT-4.5 while they adapt workflows to newer defaults. (help.openai.com) So the model lineup change is not just a naming cleanup. OpenAI is moving from a world where users picked from a shelf of public model names to a world where the default model, the fallback model, and the legacy model are all part of commercial packaging. (help.openai.com 1) (help.openai.com 2) For companies that built internal prompts, Custom GPTs, and approval workflows around specific names like GPT-4o or GPT-4.1, the hard part is no longer only performance. The hard part is migration clarity: who keeps access, for how long, in which surface, under which contract. (help.openai.com) (help.openai.com) And that is why a help-center update matters here. When the public picker changes but the application programming interface does not, OpenAI is telling customers that model names are no longer just technology labels; they are entitlements that can be bundled, retired, or extended depending on the plan. (help.openai.com) (openai.com)
