Skip to content
JehBuilds
Writing
By MD Jehad H.··4 min read·Operator playbook

GPT-5.5 retirement: what breaks in your automations

Drafted through my n8n + AI pipeline, edited by me.

October 14 is the date OpenAI pulls GPT-5.5 out of ChatGPT, ChatGPT Work, and Codex, and the businesses most exposed are the ones who forgot they ever picked that model in the first place.

OpenAI posted the change in its deprecations documentation and model release notes in mid September, around the same time it rolled out GPT-5.6 Sol, Terra, and Luna. For most people typing into ChatGPT, the switch will be invisible. Conversations just move to whatever model is default. The exposure is for anyone who built a workflow that names GPT-5.5 specifically.

The GPT-5.5 retirement, in plain terms

GPT-5.5 disappears from ChatGPT on every plan, from Free through Enterprise and Edu, plus ChatGPT Work and Codex sessions signed in with a ChatGPT account. It does not disappear from the OpenAI API. A Codex session authenticated with your own API key keeps working too, since that traffic never touches the subscription surface. The cutoff is a rule about where the model lives, not a shutdown of the model itself.

Where small businesses get caught

  • A custom GPT built for client intake or FAQ answers that was configured to run on GPT-5.5
  • A ChatGPT Team or Enterprise workspace with GPT-5.5 set as the default model for the whole team
  • A scheduled task in ChatGPT, like a weekly report or daily summary, that references GPT-5.5 by name
  • A Codex script signed in with your ChatGPT account rather than an API key
  • A Zapier, Make, or n8n step that routes through a ChatGPT-hosted custom GPT instead of the raw API

Table comparing which ChatGPT and Codex surfaces lose GPT-5.5 on October 14 and which keep it through the API

SurfaceWhat happens on October 14
ChatGPT (Free through Enterprise, Edu)GPT-5.5 retired, conversations move to the current default model
ChatGPT WorkWorkspace defaults pinned to GPT-5.5 stop responding and need a new default
Codex, signed in with ChatGPTRequests for gpt-5.5 fail, OpenAI points users to gpt-5.6-sol
Codex, signed in with an API keyUnaffected, this traffic goes through the API platform
OpenAI API (direct)Unaffected, GPT-5.5 stays available to call by ID
Only the subscription surfaces lose GPT-5.5 on October 14. Anything calling the API directly, including Codex signed in with an API key, keeps running.

What to do before October 14

  1. 1

    Find what points at GPT-5.5

    Check every custom GPT, workspace default model setting, and scheduled task for a reference to GPT-5.5. Ask whoever manages your ChatGPT Team or Enterprise workspace to pull the model settings list rather than guessing.

  2. 2

    Test your top workflow on GPT-5.6 Sol

    Run the same prompt or scheduled task manually on GPT-5.6 Sol before the forced switch happens. Compare tone, length, and accuracy so you know what changes before a client sees the difference.

  3. 3

    Repoint anything hardcoded

    Update custom GPT configurations, scheduled tasks, and Codex settings to gpt-5.6-sol or gpt-6-astra. Do it now, while you can still compare against GPT-5.5 side by side.

  4. 4

    Separate out your API traffic

    If automations call the OpenAI API directly through n8n, a Zapier code step, or a custom integration, GPT-5.5 keeps working there. That is not a reason to ignore it, just a reason to plan that migration on your own schedule instead of OpenAI's.

Do the audit this week

The failure mode here is quiet. Nobody gets an error message until October 14, when a scheduled report or a client-facing GPT just stops behaving the way it did the day before. An hour of checking now costs less than a client noticing first.

If you are not sure which of your workflows still call a model by name, that is usually the first thing worth mapping together before the next retirement catches you by surprise.

Building something this should run inside?

Book a systems call

Keep reading