GPT-5.5 leaves ChatGPT and Codex on 14 October: what changes for apps on your plan

GPT-5.5 retires from ChatGPT, ChatGPT Work and Codex on every plan on 14 October 2026, and the OpenAI API is not affected. For apps that run on your ChatGPT plan, nothing about the connection or your limits changes; you only need to move any app, Codex config or scheduled task that still names gpt-5.5 to a current model such as GPT-6 Sol before that date.

We re-read OpenAI's own pages on 9 October 2026 to confirm the date and wording. Below is what they say, which apps on BringYourSubs it touches, and a short checklist.

What exactly is OpenAI retiring on 14 October?

OpenAI's Codex pricing page carries the notice in one line: "GPT-5.5 retires from ChatGPT, ChatGPT Work, and Codex on all plans on October 14, 2026. The OpenAI API isn't affected."

The Codex models page adds the scope: the retirement covers "consumer, Business, Enterprise, and Edu plans". It first appeared in the Codex changelog on 14 September 2026, and the ChatGPT account on X announced it the next day. News coverage, such as Business Today on 16 September, repeats the same facts.

Surface	GPT-5.5 after 14 Oct 2026	Source
ChatGPT (all plans)	Retired	OpenAI pricing and models pages
ChatGPT Work	Retired	Same
Codex signed in with ChatGPT (app, CLI, IDE extension)	Retired	Same
Codex or any app using your own OpenAI API key	Still available, billed at API rates	"This retirement does not apply to the OpenAI API"

This is the second retirement of this kind in six weeks. OpenAI's workspace model availability page notes that GPT-5.4 and GPT-5.4 mini retired from Codex for ChatGPT sign-in users on 31 August 2026.

Which model replaces GPT-5.5?

OpenAI's guidance, as read on 9 October 2026:

• Plus, Pro, Business, Enterprise and Edu: choose GPT-6 Sol (gpt-6-sol) when it is available to you.
• Free and Go: choose GPT-6 Luna (gpt-6-luna) in the desktop app.
• For complex coding and agent work, the same page recommends GPT-6.1 Sol where your account and client have it, and Luna for "focused, repeatable tasks".

The advice has shifted since the announcement. The 15 September post on X pointed users to GPT-5.6 Sol or GPT-6 Astra; the docs now name GPT-6 Sol. Both GPT-5.6 Sol and Astra are still listed on the models page, so either works if you already moved to one of them.

One more naming detail matters for anyone reading the news this week. OpenAI's October system card says new versions of GPT-6 Sol and Luna went to everyone in ChatGPT, while Codex and ChatGPT Work "are still using previously released versions". The model name in your Codex picker is the same; the version behind it in Codex is the September one.

What changes for apps that run on your ChatGPT plan?

Apps on BringYourSubs use your ChatGPT plan in one of two ways, and the retirement reaches both.

1. Apps that drive the Codex you signed in to. Conductor, T3 Code, Happy, Munder Difflin, Superset, CloudCLI and goose start the official Codex CLI or app-server on your machine. They are Codex with a different front end, so OpenAI's Codex instructions apply directly: wherever a model is chosen, gpt-5.5 has to go.

2. Apps that connect through Sign in with ChatGPT. OpenAI's partner page lists Amp, Conductor, Dactyl, Devin, Hermes Agent, Hyperagent, Kilo Code, Notion, Vercel (v0), Vorflux and Warp, plus the open-source OpenClaw, OpenCode, Pi and T3 Code. These apps send requests on your plan's Codex and ChatGPT Work allowance.

OpenAI's pages do not describe the Sign in with ChatGPT route separately in the retirement notice. What they do say is that plan usage "uses your Codex / ChatGPT work usage included in your plan", and the developer docs tell apps to build their model picker from the signed-in account's own model list and to "use a model available to the signed-in account". Our reading: once GPT-5.5 is gone from your plan, an app that follows those docs simply stops offering it.

Some apps already work that way. Kilo's docs say it "uses OpenAI's available-model list to determine eligible models". OpenCode's Sign in with ChatGPT method, merged on 29 September, caches the account's model list and refreshes it every 15 minutes. The risk sits with apps, agents or scripts that saved gpt-5.5 by name.

What does not change: the connection itself, your plan's limits, any weekly cap you set per app under ChatGPT Settings, Usage, and the fact that the app cannot see your chats. Apps on a Claude or Cursor plan are not affected at all.

Will switching models use more of your plan?

Not on the credit rate card. OpenAI's pricing page lists these Standard-speed rates per million tokens (read 9 October 2026):

Model	Input	Cached input	Output
GPT-5.5	125 credits	12.5 credits	750 credits
GPT-6.1 Sol	50 credits	2.5 credits	250 credits
GPT-6 Sol	50 credits	5 credits	250 credits
GPT-6 Luna	2.5 credits	0.25 credits	12.5 credits

OpenAI adds a caveat: "Credit prices alone don't determine included subscription usage." Treat the table as a guide to relative cost, and check your usage dashboard (/status in the Codex CLI) after a few days on the new model.

What should you check before 14 October?

OpenAI's own list of places to look is specific: "workspace defaults, saved model settings, managed configurations, custom agents, scheduled tasks, and scripts that select a model". In practice, for a person running apps on their own plan:

1. Codex config. Open your Codex settings and any model line in your Codex config file. If it says gpt-5.5, change it. In the CLI, /model switches the current session.
2. Each app's model picker. Open Conductor, OpenCode (/models), Kilo Code, Hermes Agent and any other app you use on your plan, and pick a current model.
3. Scheduled and headless runs. Scripts that call codex exec -m gpt-5.5, cron jobs and custom agents fail quietly at night. Search your repos and dotfiles for gpt-5.5.
4. Team workspaces. On Business, Enterprise and Edu, an admin must enable the replacement first. OpenAI's admin page notes that "changing a default doesn't grant model access."
5. If you really need GPT-5.5 after the 14th, the only route is an OpenAI API key, billed per token. Some apps accept a key next to the plan sign-in; that usage does not count against your plan.

Where can you see every app that runs on your ChatGPT plan?

The full list, with how each one connects, is on works with ChatGPT. If you are new to the idea, our explainer on using your own Claude or ChatGPT subscription in other apps covers the cost and the rules, and the step-by-step plan connection guide shows how to sign Codex in to your plan. For OpenCode specifically, see OpenCode on your ChatGPT plan.