OpenAI drops GPT-5.5 from ChatGPT on 14 October
Reviewed by Helen Jones
OpenAI said on 15 September 2026 that GPT-5.5 will leave ChatGPT, ChatGPT Work, and Codex on 14 October. The cutoff covers all ChatGPT plans. Codex users signed in with ChatGPT were told to move to GPT-5.6 Sol or GPT-6 Astra. The OpenAI API and Codex sessions that use an API key can still call GPT-5.5. Saved model choices, workspace defaults, custom agents, and scheduled tasks that pin GPT-5.5 in the ChatGPT products will not. Ordinary chats will follow whatever default OpenAI sets after the date. GPT-5.5 launched on 23 April 2026. As a named ChatGPT frontier model it lasted under six months.
Until now, many teams treated a named model as a stable workhorse. They locked GPT-5.5 into Projects, custom GPTs, and repeating jobs so the output would not drift. That felt like a system. It was really a short lease. After 14 October, those pins fail in the products people actually use, even if the API still has the old model. The useful change is not a new slogan about agents. It is that reusable workflows built on a specific ChatGPT model name now have a hard expiry, and the replacement will not match tone, length, or error patterns unless you re-check the work.
Analysis
This is a change to act on, not a reason to rebuild your whole stack. Open every Project, custom GPT, scheduled task, and team default that still names GPT-5.5, switch it to the successor your plan offers, and run one real recurring job through the same review step you already use so you catch tone and fact drift before 14 October.
Source note
Pulse published by Collab365 Spaces, reviewed by Helen Jones on . Cite as "OpenAI drops GPT-5.5 from ChatGPT on 14 October", Collab365 Spaces. 2 sources referenced.