Service eligibility and regional restrictions
PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.
Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.
Key Takeaways
- The announcement covers ChatGPT, ChatGPT Work and Codex for consumer, Business, Enterprise and Edu plans. The OpenAI API is not affected by this GPT-5.5 retirement. The exact switchover time and time zone have not been announced; a plan is not an implemented change.
- For Codex with ChatGPT sign-in, OpenAI recommends changing explicit gpt-5.5 selections to gpt-5.6-sol before October 14. With API-key authentication, check model availability separately for the relevant API organization and project.
- Inventory workspace defaults, saved model selections, managed configuration, custom agents, scheduled tasks and scripts that explicitly choose a model. Changing one default neither updates every other entry point automatically nor grants access to the replacement.
- Confirm replacement availability for affected members and clients, then test output quality, tool behavior, duration and cost with representative tasks. Different models cannot be assumed equivalent.
- Model retirement concerns product and authentication boundaries, not a network outage. PRODUCT_FIT=NONE: changing IP addresses or proxies cannot restore a retired model.
First establish which GPT-5.5 context is involved
The Models and Workspace model availability pages in OpenAI Docs both state that GPT-5.5 will leave ChatGPT, ChatGPT Work and Codex on October 14, 2026, across consumer, Business, Enterprise and Edu plans. This retirement does not apply to the OpenAI API. No exact cutoff time or time zone was specified. On September 19 it should still be described as an announced future plan, not an already completed removal.
The same person may select models in ChatGPT, sign into the Codex desktop app or CLI with ChatGPT, or use an API key. Product entry points and authentication boundaries differ. Workspace ChatGPT model settings do not automatically apply to Codex CLI, IDE, cloud or API access. Do not infer permissions at other entry points from one model list.
Who needs to act, and who should not be pushed into an unsupported API migration
People manually using GPT-5.5 in ChatGPT or ChatGPT Work should check available replacements and differences from established workflows. Teams using ChatGPT sign-in for Codex with gpt-5.5 pinned in configuration, agents, scheduled tasks or scripts should especially complete replacement and regression testing before the date. Forgotten explicit selections may cause subsequent runs to fail or not start as expected.
Callers using this model only through the OpenAI API are outside this product-side retirement. Do not tell API teams to change model IDs without a basis in this announcement. API availability still depends on the current official model catalog and organization and project permissions; this announcement does not guarantee that the API will remain unchanged forever.
Build an inventory of pinned-model entry points
For each product surface in use, record its owner, authentication method, current model ID, affected members, execution frequency and test results. Codex users signing in with ChatGPT should at least inspect workspace defaults, saved model settings, managed configuration, custom agents, scheduled tasks and explicit gpt-5.5 selections in scripts or command lines. ChatGPT chat defaults may be independent of Work/Codex defaults; changing one does not establish completion everywhere.
If scripts contain an old-model fallback chain, automation runs on a schedule or managed policies are shared by several people, record the editor, reviewer and planned effective time too. This describes the inspection scope; it does not access readers’ workspaces or offer configuration techniques to bypass organizational policy.
A recommended replacement still requires client-by-client eligibility checks
For Codex with ChatGPT sign-in, OpenAI explicitly recommends replacing gpt-5.5 with gpt-5.6-sol. Setting a replacement default as an administrator does not grant model access. Seats, roles, workspace switches, client support and rollout timing may still make it unavailable. Verify separately in the ChatGPT, desktop, CLI, IDE or cloud clients the target users actually use before changing dependent tasks.
Select familiar, reproducible tasks and compare output, tool calls, file changes, duration and billing or usage boundaries. GPT-5.6 Sol is the official migration recommendation, not a promise of identical performance, cost or behavior in every scenario. Have administrators review managed configuration through established organizational procedures.
When validation fails, stay at the permissions and configuration layer
If the replacement cannot be selected, first check the product surface, ChatGPT sign-in versus API key, workspace or API project, user role and client support, then whether the relevant settings have taken effect. A failed automated task does not establish that the October 14 retirement happened early; model permissions, configuration syntax or a problem in the task itself may be responsible.
Before the planned date, retain traceable old configuration and representative output for change review. Reversible local configuration changes may be restored and the switch paused under organizational procedures, but reverting to gpt-5.5 after retirement takes effect will not make it available again. Work that needs more time should find a supported alternative in advance, not rely on changing network egress.
Recheck conditions and product boundaries
As of September 19, 2026, official material confirms the retirement date, product scope, API exclusion and replacement recommendation for Codex with ChatGPT sign-in. The actual cutoff time, complete permissions of individual accounts and execution results remain unknown. Recheck OpenAI Docs, workspace model availability and in-product notices near October 14; this article is not a permanent model catalog.
PRODUCT_FIT=NONE: a retired or unauthorized model is a product availability and account permissions issue. A fixed IP, proxy or region change cannot grant model access or replace configuration and workflow testing.
Sources
Frequently Asked Questions
Will GPT-5.5 also retire from the OpenAI API on October 14?
This announcement does not apply to the OpenAI API. API callers should separately check the official API model catalog and organization and project permissions, not mechanically apply the ChatGPT/Codex retirement scope to the API.
What should replace gpt-5.5 in Codex with ChatGPT sign-in?
OpenAI recommends gpt-5.6-sol. Inspect defaults, saved settings, managed configuration, agents, scheduled tasks and scripts. Verify actual availability for each member and client.
Does an administrator changing the workspace default migrate every task automatically?
Do not assume so. Tasks, agents, scripts and saved personal settings explicitly pinned to the old model still need an inventory. Changing a default also does not grant additional access to the replacement model.
Can changing network egress keep a retired model selectable?
No. Model availability and permissions depend on the product, sign-in method and workspace or API project. A proxy or fixed IP does not change the retirement plan or replace migration validation.