The classic big language model that developers heavily rely on is ushering in a platform dominated forced retirement trend. The official team of Codex recently announced that starting from June 2, 2026, multiple stable old models such as GPT-5.2 and GPT-5.3-Cdex will be fully discontinued, and a new flagship GPT-5.5 will be promoted. This move quickly sparked a fierce backlash in the developer community, with the focus of controversy pointing directly to the "intelligence reduction" and experience regression of the new model.
Many developers openly complained about GPT-5.5. GPT-5.5 showed significant regression in actual production:logical reasoning ability declined, response speed slowed, and complex problem processing time significantly increased. To ensure business stability, a large number of developers have voluntarily rolled back to more mature GPT-5.3-Cdex or GPT-5.4 to avoid the instability risks of the new version.
Now Codex is forcing the removal of old core versions, which means thatfree users will completely lose their choiceand be forced to bind to the controversial GPT-5.5.; Small teams and individual developers who rely on the stability of older versions face dual pressures of business adaptation and stability.
Faced with community protests, Codex has launched alayered exemption schemeto appease dissatisfaction:
This forced retirement is essentially OpenAI's strategic action to promotemodel ecological unification- since GPT-5.4, the Codex dedicated programming branch has gradually been integrated into the main model, and GPT-5.3 has become the last generation of independent Codex models.. The platform hopes to optimize GPT-5.5 by streamlining the version matrix and concentrating resources, but the aggressive forced replacement strategy ignores the core demands of developers for stability, speed, and reliability, ultimately leading to a trust crisis.