CONFIRMED: GPT-5.5 leaves the consumer product, not the infrastructure

OpenAI set October 14 as the farewell date for GPT-5.5 in ChatGPT, ChatGPT Work, and Codex used with plan authentication. The change was announced by the official ChatGPT account and clarified by OpenAI Developers: the model will remain available on the API platform and in Codex sessions authenticated with an API key. This is a surface retirement, not a complete shutdown of the model.

The same version now has two very different destinations

For people using ChatGPT as a ready-made product, the consequence is direct. GPT-5.5 will disappear from the picker and from plan-based experiences on October 14. In Codex, the official guidance is to move to GPT-5.6 Sol or GPT-6 Astra. Teams calling the model through the API can keep using GPT-5.5, provided access and the identifier remain valid in their development environment. The split also applies to Codex with an API key, which does not follow the same calendar as subscription access.

What is confirmed and what is not

The October 14 deadline, the end of access in the named surfaces, and the model's continued availability through the API are confirmed. The recommendation of Sol and Astra as plan-based Codex alternatives is also confirmed. This does not mean old conversations will be migrated automatically without changes, that behavior will be identical between ChatGPT and the API, or that the endpoint will remain available forever. The posts do not detail freeze policy, feature support, or a later API sunset.

Developer accounts avoid the shock, but still need an audit

The distinction matters for products that depend on the model behind a proxy, an agent, or a custom integration. They do not need to switch immediately just because the model is leaving ChatGPT. They should still review calls pinned to older versions, limits, tools, system messages, and regression tests before October. Codex users with subscription login need to test the move to Sol or Astra separately from users with an API key. These are different authentication paths, and OpenAI is treating them as different products.

The market reading

Retiring a model only a few months after it reaches the product reinforces an uncomfortable trend: the ChatGPT picker is becoming a fast-moving product layer, while the API acts as a longer-lived compatibility zone. That gives builders more short-term predictability, but it widens the gap between the experience customers buy and the one developers integrate. For small teams, the sensible decision is not to confuse API availability with a promise of permanent stability. The real contract remains the documentation, the model actually served, and a reproducible migration test.

My reading

This announcement is less about GPT-5.5 quality than about portfolio control. OpenAI wants product users moving to newer models without forcing the entire API base to migrate on the same day. For agent operators, that is good short-term news and a reminder that subscription access and API access are not the same product. I would treat October 14 as a mandatory review date for Codex and ChatGPT workflows, not as a reason to replace an integration that still works.

Sources

ChatGPT on X, GPT-5.5 retirement announcement: https://x.com/i/status/2099954190600876533 | OpenAI Developers on X, API availability clarification: https://x.com/i/status/2099954342170067174