gpt-5.3-codex, gpt-5.1 and gpt-5.4-nano shut down on 2027-04-01: what replaces them
As of 2026-10-08, the 2026-10-01 section of OpenAI's deprecations page says gpt-5.3-codex, gpt-5.1 and gpt-5.4-nano will be removed from the API on 2027-04-01 (April 1, 2027 in the original), with six months' notice. Recommended replacements: gpt-6-sol for gpt-5.3-codex and gpt-5.1, gpt-6-luna for gpt-5.4-nano.
Updated 2026-10-08
Four numbers that matter
Shutdown date in the official table
All three rows give Apr 1, 2027 as the Shutdown date. OpenAI's definition: at shutdown the model or endpoint is no longer accessible, and a model counts as deprecated from the moment the deprecation is announced.
Models in this batch
gpt-5.3-codex, gpt-5.1 and gpt-5.4-nano. The other section posted the same day covers TTS models (removal on 2027-01-06) and is out of scope here. gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini and gpt-5.1-chat-latest are not in this batch either: the official table says they were shut down on 2026-07-23.
Notice period for this batch
The original says with six months' notice. OpenAI's policy promises at least 6 months for generally available models and at least 3 months for specialized variants, naming gpt-5.3-codex as a Codex variant; this batch got 6 months.
Default reasoning.effort (old → new)
Per the official model pages, gpt-5.1 and gpt-5.4-nano default to none, while gpt-6-sol and gpt-6-luna support none through max and default to medium. Swap only the model and the reasoning level changes with it.
What is being shut down
As of 2026-10-08, the answer is: OpenAI will remove gpt-5.3-codex, gpt-5.1 and gpt-5.4-nano from the API on 2027-04-01. The 2026-10-01 section of the deprecations page states that these models are deprecated and will be removed from the API on April 1, 2027, with six months' notice, and asks developers to migrate to the recommended replacements before the shutdown date. OpenAI says it emails customers who are actively using a deprecated model and documents the deprecation on that page; as of 2026-10-08 this section sits at the top of Upcoming deprecations. Keep two other timelines apart: OpenAI's Codex docs say gpt-5.3-codex is already deprecated in Codex when you sign in with ChatGPT, and on the QCode side the gpt-5.3-codex ID is no longer available on QCode; available models are listed on /models.
The three official rows (checked 2026-10-08)
Every row has the shutdown date Apr 1, 2027: ① gpt-5.3-codex → gpt-6-sol; ② gpt-5.4-nano → gpt-6-luna; ③ gpt-5.1 → gpt-6-sol. The columns are Shutdown date, Model / system and Recommended replacement; the replacement column is copied verbatim. The table lists undated model IDs: the model pages for gpt-5.1 and gpt-5.4-nano give gpt-5.1-2025-11-13 and gpt-5.4-nano-2026-03-17 as the Default snapshot, and the table does not list those dated snapshots as separate rows; for gpt-5.3-codex the Default snapshot is gpt-5.3-codex itself. The gpt-6-sol model page adds: See GPT-6.1 Sol for the newer Sol model.
Timeline
OpenAI adds this batch to its deprecations page: gpt-5.3-codex, gpt-5.1 and gpt-5.4-nano are deprecated and will be removed from the API on 2027-04-01. A separate section the same day schedules TTS models for removal on 2027-01-06 (not covered here).
We re-fetched the deprecations page and the model pages: the three rows are still under Upcoming deprecations, the date is still Apr 1, 2027 and the replacements are unchanged.
Shutdown: the three models are removed from the API and, per OpenAI's definition, are no longer accessible. OpenAI also notes that in some cases developers may be able to provision dedicated capacity through its sales team for continued access after the shutdown date.
Confirmed vs not stated by OpenAI
Confirmed (verbatim on official pages)
All of the following can be checked word for word on official pages fetched on 2026-10-08: the 2026-10-01 section, removal on April 1, 2027 and with six months' notice; the three model and replacement rows; each model page's Model ID and Default snapshot; OpenAI's definition Snapshots let you lock in a specific version of the model so that performance and behavior remain consistent; gpt-5.3-codex supports Responses but not Chat Completions, with a maximum input of 272,000; gpt-6-sol and gpt-6-luna have a 1,050,000 context window and a 922,000 maximum input; the Codex docs say gpt-5.2 and gpt-5.3-codex are already deprecated in Codex when you sign in with ChatGPT.
⚠️ Not stated by OpenAI / watch out
① Whether the dated snapshots gpt-5.1-2025-11-13 and gpt-5.4-nano-2026-03-17 go on the same day is not spelled out in the table; they are the only snapshots on those model pages, so we suggest treating them as affected (our reading, not OpenAI's). ② Whether IDs like gpt-5.1 will return errors or be pointed at a new model after shutdown is not stated; do not count on automatic switching. ③ The table does not promise equivalent output or equal prices. ④ The Codex docs do not say when gpt-5.3-codex stops in Codex signed in with your own API key; the API table gives 2027-04-01. ⑤ As of 2026-10-08 there is no change of date.
Direct API / SDK users vs Codex users
Calling the API or SDK directly
The model string is not the only change. From the official model pages: gpt-5.1 and gpt-5.4-nano default to reasoning.effort none, gpt-6-sol and gpt-6-luna default to medium, so set effort explicitly to keep the old behavior; on Chat Completions, gpt-6-sol supports function calling only with reasoning_effort set to none, so tool-using code is better moved to Responses; gpt-5.3-codex was Responses-only already. Context grows from 400,000 to 1,050,000, but a prompt with more than 272K input tokens is priced at 2x input and cache rates and 1.5x output for the full request (OpenAI's wording).
Using Codex
The model lives in the model entry of config.toml; also check codex exec --model or -m in scripts (OpenAI's example is codex exec -m gpt-6.1-sol). For Codex signed in with ChatGPT, OpenAI says gpt-5.3-codex is already deprecated and asks you to update scripts, configuration files and codex exec --model commands. With a custom model provider or gateway, use a compatible Responses API endpoint; current Codex releases don't support wire_api = "chat" or a Chat Completions-only endpoint.
Migration checklist (before 2027-04-01)
① Search for the IDs: look for gpt-5.3-codex, gpt-5.1 and gpt-5.4-nano, plus the dated gpt-5.1-2025-11-13 and gpt-5.4-nano-2026-03-17, in code, environment variables, config files, eval scripts and CI. ② Match whole words: when searching for gpt-5.1, exclude gpt-5.1-codex and gpt-5.1-chat-latest (the 2026-07-23 batch, already shut down); when searching for gpt-5.4-nano, include the dated form. ③ Tell snapshots from undated IDs: OpenAI defines a snapshot as a way to lock in a specific version of the model so that performance and behavior remain consistent, while an alias points requests at a model, as in OpenAI's sentence The gpt-5.6 alias routes requests to GPT-5.6 Sol. We suggest changing both forms before shutdown. ④ Swap per the table: gpt-5.3-codex and gpt-5.1 → gpt-6-sol, gpt-5.4-nano → gpt-6-luna; the gpt-6-sol page also points to the newer gpt-6.1-sol, which does not support the none and minimal effort levels. ⑤ Check parameters: set reasoning.effort explicitly; move tool-using Chat Completions calls to Responses; budget long-context pricing for inputs that may exceed 272K. ⑥ Verify: send a batch of real requests to the endpoint you actually use and confirm that the model in responses and billing logs is the new ID.
Using the replacements on QCode
As of 2026-10-08, the gpt-5.3-codex ID is no longer available on QCode; available models are listed on /models. Of the official replacements, gpt-6-sol is callable on QCode, and so is gpt-6.1-sol, the newer Sol the gpt-6-sol page points to; gpt-6-luna is callable too, with its price shown on /models. This page does not cover the status of gpt-5.1 or gpt-5.4-nano on our service; see /models. GPT models use the OpenAI endpoints, not the Anthropic endpoint: in the OpenAI SDK set base_url to https://api.qcode.cc/openai/v1 (requests go to /openai/v1/chat/completions); in Codex CLI set base_url = https://api.qcode.cc/openai in the TOML config with wire_api set to responses (requests go to /openai/v1/responses). The Codex tutorial in the docs recommends gpt-6-sol as the default model. Billing is per token; see /models for each model's price. GPT series: when a single request's input (including cached tokens) exceeds 272K tokens, the whole request is billed at long-context rates, with input and cache-read prices ×2 and output ×1.5, matching OpenAI's official rule; see the >272K row under each model's price on /models.
Frequently asked questions
When are gpt-5.3-codex and gpt-5.1 shut down?
On 2027-04-01, when they are removed from the OpenAI API. As of 2026-10-08, the 2026-10-01 section of OpenAI's deprecations page lists Apr 1, 2027 as the shutdown date for all three rows (gpt-5.3-codex, gpt-5.1, gpt-5.4-nano), with six months' notice. Per OpenAI's definition, a model is no longer accessible at shutdown.
Why do some sources say gpt-5.3-codex is already gone?
Because they describe different places. 2027-04-01 is the removal date from the official OpenAI API, per the deprecations page; OpenAI's Codex docs separately say gpt-5.3-codex is already deprecated in Codex when you sign in with ChatGPT; and on the QCode side the gpt-5.3-codex ID is no longer available on QCode; available models are listed on /models.
What changes when moving from gpt-5.1 to gpt-6-sol?
Check at least three things: change the model to gpt-6-sol; set reasoning.effort explicitly, because gpt-5.1 defaults to none and gpt-6-sol to medium; and if you use function calling on Chat Completions, gpt-6-sol supports it only with reasoning_effort set to none, so tool-using calls are better moved to Responses. The gpt-6-sol page also points to the newer gpt-6.1-sol, which does not support none or minimal and supports Chat Completions only without tool calling.
What replaces gpt-5.4-nano?
The official table maps gpt-5.4-nano → gpt-6-luna. OpenAI calls GPT-6 Luna its most efficient model for focused, high-volume tasks; the default effort moves from none to medium, so set it explicitly to keep the old behavior. gpt-6-luna is callable on QCode with its price shown on /models; gpt-5.6-luna has been removed from /models and is not recommended for new integrations.
What do Codex users need to change?
The model entry in config.toml and the model after codex exec --model (or -m) in scripts. OpenAI's instruction is to add a model entry to your configuration file; for Codex signed in with ChatGPT, OpenAI says gpt-5.3-codex is already deprecated and asks you to update scripts, configuration files and codex exec --model commands. With a custom model provider or gateway, use a compatible Responses API endpoint; current Codex releases don't support wire_api = "chat".
Can I call the official replacements through QCode?
Yes: as of 2026-10-08, gpt-6-sol and gpt-6.1-sol are callable on QCode, and gpt-6-luna is too, with its price shown on /models; billing is per token, see /models for each model's price. In Codex set base_url = https://api.qcode.cc/openai in the TOML config; in the OpenAI SDK use https://api.qcode.cc/openai/v1.
Sources
Shutdown date, models and replacements: OpenAI's official deprecations page https://developers.openai.com/api/docs/deprecations (and its .md version), fetched 2026-10-08, using the 2026-10-01 section, the notice-period policy and the deprecation definitions, cross-checked against the 2026-07-23 section of the same page. Model ID, Default snapshot, the snapshot definition, context and maximum input, supported reasoning.effort values and endpoint support: OpenAI's model pages for gpt-5.1, gpt-5.3-codex, gpt-5.4-nano, gpt-6-sol, gpt-6-luna, gpt-6.1-sol and gpt-5.6-sol plus the models overview, fetched the same day. Codex: the Models page of OpenAI's Codex docs (learn.chatgpt.com), same day. QCode setup: the docs.qcode.cc pages on endpoints and API formats and the Codex tutorial (fetched 2026-10-08). OpenAI writes dates as Apr 1, 2027; this page uses ISO format.
Replace the three old IDs before 2027-04-01
As of 2026-10-08, the official replacement gpt-6-sol and its newer version gpt-6.1-sol are callable on QCode, billed per token; see /models for each model's price. Availability of old IDs follows /models.
Related reading
GPT-5.3-Codex is retired: migrating to GPT-5.6
Our page on gpt-5.3-codex being disabled on QCode and how to migrate; a different matter from the OpenAI API date on this page.
gpt-5 and o3 snapshots shut down on 2026-12-11
The previous batch: which model replaces each of the six GPT-5 / o3 snapshots and how to handle aliases.
GPT-6.1 Sol API guide
The newer Sol the gpt-6-sol page points to: pricing, the 272K rule and how it differs.
Not affiliated with OpenAI. Dates, models and replacement mappings on this page are taken from OpenAI's official deprecations page, model pages and Codex docs fetched on 2026-10-08; the official pages prevail. Model availability is as listed on /models; upstream changes may happen without notice. This is not OpenAI's migration guide and makes no promise that any old ID stays available after its shutdown date.