GPT-5 and o3 snapshots shut down on 2026-12-11: what replaces the 6 IDs
As of 2026-10-08, the 2026-06-11 section of OpenAI's deprecations page still says six snapshots leave the API on 2026-12-11 (Dec 11, 2026 in the original): gpt-5-2025-08-07, gpt-5-mini-2025-08-07, gpt-5-nano-2025-08-07, gpt-5-pro-2025-10-06, o3-2025-04-16 and o3-pro-2025-06-10. Official replacements: gpt-5 and o3 → gpt-5.6-sol, gpt-5-pro and o3-pro → gpt-5.6-sol with reasoning.mode: pro, gpt-5-mini → gpt-5.6-terra, gpt-5-nano → gpt-5.6-luna (gpt-5.6-luna has been removed from QCode /models and is not recommended for new integrations).
Updated 2026-10-08
Four numbers that matter
Shutdown date in the official table
The table columns are Shutdown date, Model / system and Recommended replacement; all six rows read Dec 11, 2026. Official definition: at the time of the shutdown, the model or endpoint will no longer be accessible.
Snapshots in this batch
One dated snapshot each for gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3 and o3-pro, i.e. the Default snapshot on each model page. gpt-5-codex and gpt-5-chat-latest are not in this batch: the official table lists them as shut down on 2026-07-23.
Rows landing on sol / terra / luna
Four rows go to gpt-5.6-sol (gpt-5-pro and o3-pro with reasoning.mode: pro), gpt-5-mini goes to gpt-5.6-terra and gpt-5-nano to gpt-5.6-luna. OpenAI's model pages say Sol, Terra and Luna roughly correspond to the unsuffixed, mini and nano tiers of earlier GPT-5 families. gpt-5.6-luna has been removed from QCode /models and is not recommended for new integrations.
Max input tokens (gpt-5 → gpt-5.6-sol)
Official model pages: gpt-5 has a 400,000 context window and 272,000 max input tokens; gpt-5.6-sol has 1,050,000 and 922,000. The gpt-5.6-sol page adds that prompts with >272K input tokens are priced at 2x input and 1.5x output for the full request.
What is being retired
As of 2026-10-08, the answer is: OpenAI will remove six older GPT-5 and o3 snapshots from the API on 2026-12-11. The 2026-06-11 section of the deprecations page says: "On June 11, 2026, we notified developers using older GPT-5 and o3 model snapshots of their deprecation and removal from the API on December 11, 2026." Each of the six IDs is the Default snapshot on the model page for gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3 or o3-pro, and it is the only entry in that page's Snapshots list. OpenAI promises at least 6 months' notice for generally available models; June 11 to December 11 is exactly six months by the calendar. As of 2026-10-08 the section still sits under the Upcoming deprecations heading.
The six official rows (each checked 2026-10-08)
Every row has the shutdown date Dec 11, 2026: ① gpt-5-2025-08-07 → gpt-5.6-sol; ② gpt-5-mini-2025-08-07 → gpt-5.6-terra; ③ gpt-5-nano-2025-08-07 → gpt-5.6-luna; ④ gpt-5-pro-2025-10-06 → gpt-5.6-sol (reasoning.mode: pro); ⑤ o3-2025-04-16 → gpt-5.6-sol; ⑥ o3-pro-2025-06-10 → gpt-5.6-sol (reasoning.mode: pro). All six rows were re-fetched and checked one by one on 2026-10-08; the replacement column is copied verbatim. Unlike the 2026-10-23 batch, each cell lists only the dated snapshot and no alias such as gpt-5 or o3. Note: gpt-5.6-luna has been removed from QCode /models and is not recommended for new integrations.
Timeline
OpenAI notifies developers using older GPT-5 and o3 snapshots that these six snapshots are deprecated and will be removed from the API on 2026-12-11 (see the 2026-06-11 section of the deprecations page).
We re-fetch the official deprecations page: all six rows are still under Upcoming deprecations, the date is still Dec 11, 2026 and the replacements are unchanged. The 2026-10-23 batch (o1, o3-mini, o4-mini and others) has its own page.
Shutdown date: the six snapshots leave the API. By OpenAI's definition the model is no longer accessible at that point; OpenAI also notes that in some cases developers may be able to provision dedicated capacity, via its sales team, for continued access after the shutdown date.
Confirmed vs not stated by OpenAI
Confirmed (verbatim on official pages)
All of this can be read verbatim on the official pages fetched on 2026-10-08: the 2026-06-11 notice and the Dec 11, 2026 shutdown; the six snapshot IDs and their replacements, including reasoning.mode: pro on two rows; the Model ID and Default snapshot on each model page; the official definition "Snapshots let you lock in a specific version of the model so that performance and behavior remain consistent"; GPT-5.6 and GPT-6 models support standard and pro reasoning modes in the Responses API, with standard as the default; gpt-5-pro and o3-pro are available in the Responses API only; gpt-5-pro only supports reasoning.effort: high.
⚠️ Not stated / watch out
① Whether the aliases gpt-5, gpt-5-mini, o3 and so on stop on 2026-12-11 is not stated: the table lists dated snapshots only, and the alias list at the end of the page has no entry for them. Their only snapshot on each model page is the one being removed, so we suggest treating them as affected (our suggestion, not OpenAI's wording). ② Whether an alias will be pointed at a newer model is not stated either; do not count on it. ③ The table promises neither equivalent output nor equal pricing. ④ As of 2026-10-08 there is no rescheduling wording. ⑤ We have not verified how reasoning.mode: pro behaves on QCode. ⑥ gpt-5.1 is not in this batch; OpenAI lists it for 2027-04-01 with gpt-6-sol as the replacement.
API users vs Codex and third-party tool users
Calling the API or an SDK directly
More than the model string changes. Per the official model pages, gpt-5 supports reasoning.effort minimal, low, medium and high, while gpt-5.6-sol supports none, low, medium (default), high, xhigh and max, so minimal is not on the new list. gpt-5 takes up to 272,000 input tokens and gpt-5.6-sol up to 922,000; requests above 272K are billed at the long-context price for the whole request. Official prices: gpt-5 $1.25 input / $10 output per million tokens, gpt-5.6-sol $4 / $20 (marked as promotional pricing, available at least through 2026-11-21).
Using Codex or a third-party tool
In Codex the model lives in the model entry of config.toml (official wording: To specify a model, add a model entry to your configuration file); also check codex exec --model in scripts. gpt-5-codex is not in this batch: the official table lists it as shut down on 2026-07-23 with gpt-5.6-sol as the replacement. A custom provider or gateway must be a compatible Responses API endpoint; current Codex releases don't support wire_api = "chat". If you typed gpt-5 or o3 into a model dropdown or custom model field in tools such as Cursor or Cline, change it by the same table.
Migration checklist (before 2026-12-11)
① Search snapshots: grep code, environment variables, config files and eval scripts for gpt-5-2025-08-07, gpt-5-mini-2025-08-07, gpt-5-nano-2025-08-07, gpt-5-pro-2025-10-06, o3-2025-04-16 and o3-pro-2025-06-10. ② Search aliases: also search the quoted strings "gpt-5", "gpt-5-mini", "gpt-5-nano", "gpt-5-pro", "o3" and "o3-pro" as whole words so that o3 does not match o3-mini. OpenAI says snapshots "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 the official sentence "The gpt-5.6 alias routes requests to GPT-5.6 Sol". Each alias above has the to-be-removed snapshot as its Default snapshot and OpenAI does not say what happens to it, so we suggest changing the aliases too. ③ Replace per row: gpt-5 and o3 → gpt-5.6-sol, gpt-5-mini → gpt-5.6-terra, gpt-5-nano → gpt-5.6-luna (removed from QCode /models, not recommended for new integrations). ④ Pro rows: set model to gpt-5.6-sol and add reasoning.mode: pro, which OpenAI documents for the Responses API. ⑤ Check parameters: replace reasoning.effort minimal with a value the new model supports, and budget for the long-context price if inputs can exceed 272K. ⑥ Verify: send a batch of real requests to the endpoint you actually use and confirm that the model in responses and usage logs is the new ID.
Using the replacements on QCode
As of 2026-10-08, the official replacements gpt-5.6-sol and gpt-5.6-terra are callable on QCode. The gpt-5 model page also recommends GPT-6 Astra; gpt-6-astra is callable on QCode too. gpt-5.6-luna has been removed from /models and is not recommended for new integrations; if you used gpt-5-nano, consider gpt-6-luna, which OpenAI's model overview lists for cost-sensitive, high-volume workloads (our suggestion, not the official replacement for that row; see /models for its price). GPT models use the OpenAI endpoints, not the Anthropic endpoint: set the OpenAI SDK base_url to https://api.qcode.cc/openai/v1 (requests go to /openai/v1/chat/completions), and in the Codex CLI TOML config set base_url = https://api.qcode.cc/openai (requests go to /openai/v1/responses). Billing is per token; see /models for each model's price. For GPT models, when a single request's input (including cached tokens) exceeds 272K tokens, the whole request is billed at the long-context price: input and cache-read prices ×2, output ×1.5, the same rule as OpenAI's; each model's rate is in the >272K row under its price on /models. The status of the old snapshots on this site is not covered here; /models is authoritative.
Frequently asked questions
When are gpt-5 and o3 shut down?
On 2026-12-11. As of 2026-10-08, OpenAI's deprecations page gives six rows the shutdown date Dec 11, 2026: gpt-5-2025-08-07, o3-2025-04-16 and the matching snapshots of gpt-5-mini, gpt-5-nano, gpt-5-pro and o3-pro. Developers were notified on 2026-06-11. By the official definition, the model is no longer accessible at shutdown.
My code only says gpt-5, without a date. Am I affected?
We suggest treating it as affected; OpenAI does not say whether aliases stop on the same day. This table lists dated snapshots only, unlike the 2026-10-23 batch, where o1-2024-12-17 and the alias o1 share one cell. The gpt-5 model page names gpt-5-2025-08-07 as the Default snapshot, and it is the only one in its Snapshots list. OpenAI does not say whether the alias will error or point elsewhere after shutdown, so the safe move is to change it before 2026-12-11.
What changes for gpt-5-pro and o3-pro?
At least two things: set model to gpt-5.6-sol and add reasoning.mode: pro. OpenAI says GPT-5.6 and GPT-6 models support standard and pro reasoning modes in the Responses API and that standard is the default, so swapping only the ID drops you to standard. gpt-5-pro and o3-pro were already Responses API only; gpt-5-pro only supports reasoning.effort: high, while gpt-5.6-sol defaults to medium, so set high explicitly if you need it.
What replaces gpt-5-mini and gpt-5-nano?
Per the official table, gpt-5-mini-2025-08-07 → gpt-5.6-terra and gpt-5-nano-2025-08-07 → gpt-5.6-luna. OpenAI's model pages say Terra roughly corresponds to the mini tier and Luna to the nano tier of earlier GPT-5 families. Note that gpt-5.6-luna has been removed from QCode /models and is not recommended for new integrations; here you can look at gpt-6-luna instead (our suggestion, not an official mapping), priced as shown on /models.
What do Codex users need to change?
The model entry in config.toml and any model named after codex exec --model in scripts. OpenAI's wording: To specify a model, add a model entry to your configuration file. gpt-5-codex is not in this batch; the official table lists it as shut down on 2026-07-23, replaced by gpt-5.6-sol. With a custom provider, use a compatible Responses API endpoint; current Codex releases don't support a Chat Completions-only endpoint.
Can I call the official replacements on QCode?
Yes for two of them: as of 2026-10-08, gpt-5.6-sol and gpt-5.6-terra from the official table are callable on QCode, billed per token with prices on /models. gpt-5.6-luna is not: it has been removed from /models and is not recommended for new integrations. Whether the old snapshots can still be called after the shutdown date is not promised here; /models is authoritative.
Sources
Shutdown dates, snapshot IDs and replacements: OpenAI's official deprecations page https://developers.openai.com/api/docs/deprecations (and its .md version), fetched 2026-10-08, using the 2026-06-11 section, the notice-period policy and the deprecation definition, cross-checked against the 2026-10-23, 2026-07-23 and 2027-04-01 sections of the same page. Model ID, Default snapshot, the snapshot definition, context and max input, supported reasoning.effort values and prices: OpenAI's model pages for gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3, o3-pro and GPT-5.6 Sol / Terra / Luna plus the models overview, fetched the same day. reasoning.mode: OpenAI's reasoning models guide. Codex configuration: the Models and Config basics pages of OpenAI's Codex docs. QCode setup: the docs.qcode.cc page on endpoints and API paths (fetched 2026-10-08). Official dates such as Dec 11, 2026 are written in ISO format on this page.
Replace the six old snapshots before 2026-12-11
As of 2026-10-08, the replacements gpt-5.6-sol and gpt-5.6-terra are callable on QCode, billed per token, with prices on /models. We make no promise about old snapshots after their shutdown.
Related reading
OpenAI legacy model retirement dates 2026
Replacements and fixes for the 2026-09-28 and 2026-10-23 batches.
2026-10-23 image and reasoning ID shutdown
The o1, o3-mini, o4-mini and gpt-image-1 batch, where snapshots and aliases share a cell and stop together.
GPT-6.1 Sol API guide
The model OpenAI's overview suggests for balancing intelligence and cost.
Not affiliated with OpenAI. Dates, snapshot IDs and replacement mappings on this page are taken from OpenAI's official deprecations page and model pages 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.