GPT-5.3-Codex deprecated: off the OpenAI API on 2027-04-01, replacement gpt-6-sol
As of 2026-10-08, OpenAI's deprecations page says gpt-5.3-codex was deprecated on 2026-10-01 and will be removed from the API on 2027-04-01, with gpt-6-sol as the recommended replacement. QCode is a separate story: since 2026-08, calling this id on QCode returns model_not_supported. This page keeps the two timelines apart and lists the migration steps.
What happened
Keep the two timelines apart. 1) OpenAI: GPT-5.3-Codex was released in Codex on 2026-02-05. Since 2026-05-26 it has been deprecated as a user-selectable model in Codex for users signed in with ChatGPT, and OpenAI states that API-key workflows aren't affected. The 2026-10-01 deprecation notice says it will be removed from the API on 2027-04-01 (six months' notice), with gpt-6-sol as the recommended replacement. As of 2026-10-08 the official model page still lists gpt-5.3-codex: Responses endpoint only (Chat Completions not supported), 400,000-token context window, 128,000 max output tokens. 2) QCode: since 2026-08, calling gpt-5.3-codex on QCode returns model_not_supported with the message "Model 'gpt-5.3-codex' is no longer available. Please use gpt-5.6-sol / gpt-5.6-terra / gpt-5.6-luna", and as of 2026-10-08 QCode's /models does not list this id. QCode does not currently support gpt-6-luna or gpt-5.6-luna (both removed from /models); please use another GPT model, such as gpt-6.1-sol, gpt-6-sol or gpt-5.6-terra.
Official replacement: gpt-6-sol
OpenAI's deprecation table lists gpt-6-sol as the recommended replacement for gpt-5.3-codex; the same notice also schedules gpt-5.1 and gpt-5.4-nano for removal on 2027-04-01. gpt-6-sol can be called through QCode, and the same key also reaches gpt-6.1-sol, gpt-5.6-sol and gpt-5.6-terra. Billing is per token; see /models for each model's price.
How to migrate
1) Find every gpt-5.3-codex in saved configs, custom agents, CI scripts and scheduled jobs. 2) Switch the model to gpt-6-sol, the officially recommended replacement. 3) Before switching, check on OpenAI's model page which endpoints and reasoning-effort settings gpt-6-sol supports, then test on a small share of traffic. If you call the OpenAI API directly, the official deadline is 2027-04-01; if you call this id on QCode, it already returns model_not_supported, so switch now.
On QCode
As of 2026-10-08, QCode's /models does not list gpt-5.3-codex, and calling it returns model_not_supported. That is the QCode-side status, separate from the OpenAI API removal date of 2027-04-01. The recommended replacement gpt-6-sol can be called through QCode, as can gpt-6.1-sol, gpt-5.6-sol and gpt-5.6-terra; with the same key you only change the model. QCode does not currently support gpt-6-luna or gpt-5.6-luna (both removed from /models); please use another GPT model, such as gpt-6.1-sol, gpt-6-sol or gpt-5.6-terra. Billing is per token; see /models for each model's price.
FAQ
Can I still use gpt-5.3-codex?
Not on QCode: since 2026-08 calls return model_not_supported, and as of 2026-10-08 /models does not list it. On the OpenAI API, the deprecations page says it will be removed on 2027-04-01, with gpt-6-sol as the recommended replacement; in Codex, for users signed in with ChatGPT, it has been deprecated as a user-selectable model since 2026-05-26.
Which model should I migrate to?
Per OpenAI's deprecation table, the recommended replacement for gpt-5.3-codex is gpt-6-sol. On QCode, gpt-6-sol, gpt-6.1-sol, gpt-5.6-sol and gpt-5.6-terra can all be called, billed per token (see /models for prices); compare them on a small share of your own tasks before switching everything.
Do I need to change code?
Usually only the model id: SDK and keys stay the same, and QCode users switch to gpt-6-sol with the same key, no re-provisioning. On the official model page gpt-5.3-codex supports the Responses endpoint only; before switching, check on OpenAI's model page which endpoints and parameters gpt-6-sol supports.
Migrate to gpt-6-sol
gpt-6-sol, the officially recommended replacement, can be called through QCode, billed per token (see /models for prices). One key covers multiple model families.
Create an account