From OpenAI and Microsoft pages and package code, read 14 November 2026 · Updated
gpt-5, gpt-5-mini, gpt-5-nano, o3 deprecated: OpenAI's shutdown date is 11 December 2026
OpenAI removes gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3 and o3-pro from its API on 11 December 2026. Its deprecations page lists each short name in the same row as its dated snapshot, so plan for both to stop. The listed replacements are gpt-6.1-sol for gpt-5 and o3, gpt-6.1-terra for gpt-5-mini and gpt-6-luna for gpt-5-nano; for gpt-5-pro and o3-pro it is gpt-6.1-sol with reasoning.mode: pro in the Responses API.
On Azure the dates are Microsoft's, not OpenAI's: its schedule lists 19 November 2026 for o3 and o3-pro, three weeks before OpenAI's date, and dates in 2027 for the four GPT-5 models.
This week: search your code and configuration for the six names (the search is below). Then look at your libraries: eleven packages we read use one of these models as their default, so a project with no hit in its own files can still send it. Name the model yourself and test it before the date.
What OpenAI lists
Each row is from OpenAI's deprecations page, read on 14 November 2026. The page marks all six as Deprecated, with notice dated 11 June 2026.
| Model: short name, dated snapshot | Shutdown date | Listed replacement |
|---|---|---|
gpt-5, gpt-5-2025-08-07 | 11 December 2026 | gpt-6.1-sol |
gpt-5-mini, gpt-5-mini-2025-08-07 | 11 December 2026 | gpt-6.1-terra |
gpt-5-nano, gpt-5-nano-2025-08-07 | 11 December 2026 | gpt-6-luna |
gpt-5-pro, gpt-5-pro-2025-10-06 | 11 December 2026 | gpt-6.1-sol (reasoning.mode: pro) in the Responses API |
o3, o3-2025-04-16 | 11 December 2026 | gpt-6.1-sol |
o3-pro, o3-pro-2025-06-10 | 11 December 2026 | gpt-6.1-sol (reasoning.mode: pro) in the Responses API |
OpenAI's page recommends running representative requests against both models and comparing outputs, latency and cost before you switch production traffic. It also links a GPT-6 migration guide. We have not called any of the replacements.
What happens to a request after the date
OpenAI's page defines the state: "Shutdown. The model no longer accepts requests through the standard API." It does not print an error code or message for a shut-down model, and we have not seen one for these six. It says that in some cases developers may be able to provision dedicated capacity for continued access after a shutdown date, and points to its sales team.
For a sense of what a shut-down OpenAI model returns: after OpenAI shut down gpt-image-1 on 23 October, requests that named it got HTTP 400 and The model 'gpt-image-1' does not exist. (that page has the detail). That is what OpenAI returned for gpt-image-1, a different model; it is not a message we have seen for gpt-5 or o3.
You can ask the API for the date. OpenAI's API reference says GET /v1/models/{model} returns an optional shutdown_date field, described as "The date when the model will shut down, or null if not announced." The deprecations page tells you to call GET /v1/models and check each model's shutdown_date; it says the field is also on the retrieve endpoint. We have no API key to test this with, so this comes from OpenAI's documentation, not a call.
On Azure OpenAI and Microsoft Foundry
Microsoft keeps its own schedule. Its model retirement schedule page, read on 14 November 2026, lists these rows under Foundry Models sold by Azure, Azure OpenAI. The page's footer says it was last updated 23 September 2026.
| Model | Version | Lifecycle on Microsoft's page | Retirement date | Replacement on Microsoft's page |
|---|---|---|---|---|
gpt-5 | 2025-08-07 | GA | 9 February 2027 | none listed |
gpt-5-mini | 2025-08-07 | GA | 13 July 2027 | none listed |
gpt-5-nano | 2025-08-07 | GA | 9 February 2027 | none listed |
gpt-5-pro | 2025-10-06 | GA | 7 April 2027 | none listed |
o3 | 2025-04-16 | Deprecated | 19 November 2026 | gpt-5.6-sol |
o3-pro | 2025-06-10 | Deprecated | 19 November 2026 | gpt-5.6-sol |
Microsoft's lifecycle page says a model that has been retired returns 410 Gone from its API, and that its schedule page is the place to check current dates "since these details are subject to change". Microsoft's replacement names, gpt-5.6-sol in this table, are not the names OpenAI lists above. We took all of this from Microsoft's pages and did not call Azure.
On Azure you create deployments, and the clients take a deployment name: Haystack's Azure generators, for example, have an azure_deployment setting, and in haystack-ai 3.5.0 its Azure Responses generator defaults that setting to gpt-5-mini (haystack/). That is a deployment name, which you choose when you deploy, not a model ID. A search of your code finds a deployment only if its name contains one of the six names, so check the deployment list in your portal too.
Libraries that send one of these when you name no model
Each row below is from the latest release of the package we read on 14 November 2026, by unpacking the wheel from PyPI (we installed and ran nothing). Paths are relative to site-packages. Line numbers move between releases.
| Package, version read | What sends it, file:line | Sends | Set this to override |
|---|---|---|---|
| google-adk 2.14.1 | OpenAIResponsesLlm with no model, google/ | gpt-5 | model= |
| autoevals 0.4.1 | the RAGAS scorers with no model and no init(default_model=...), autoevals/ | gpt-5-nano | model= on the scorer, or init(default_model=...) |
| autoevals 0.4.1 | LLMClassifier with no model and no init(default_model=...), via autoevals/ and autoevals/ | gpt-5-mini | the same two |
| mem0ai 2.2.1 | OpenAILLM with no model in its config, mem0/ | gpt-5-mini | model in the LLM config |
| mem0ai 2.2.1 | the LLM reranker config, mem0/ | gpt-5-mini | model in the reranker config |
| opik 2.2.143 | LiteLLMChatModel with no model_name, opik/ | gpt-5-nano | model_name= |
| opik 2.2.143 | the LLM judge in suite evaluators, opik/ | gpt-5-nano | model= |
| opik 2.2.143 | the sycophancy metric, opik/ | openai/gpt-5 | model= |
| opik 2.2.143 | the default_llm setting used when a model is not provided, opik/ | openai/gpt-5-nano | set default_llm; its settings class reads environment variables with the prefix OPIK_ |
| inspect-evals 0.29.0 | the instrumentaleval task's grader, inspect_evals/ | openai/gpt-5-nano | the task's grader_model argument |
| shiny 1.9.0 | the AI test generator with --provider openai (default provider is anthropic), shiny/ | gpt-5-2025-08-07 (dated snapshot) | --model |
| vanna 2.0.2 | OpenAI integration with no model and no OPENAI_MODEL, vanna/ | gpt-5 | model= or OPENAI_MODEL |
| vanna 2.0.2 | OpenAI Responses integration, same condition, vanna/ | gpt-5 | the same two |
| geneva 0.18.0 | openai_udf with no model, geneva/ (the class field default is at :401) | gpt-5-mini | model= |
| fast-agent-mcp 0.10.54 | OpenAI provider when neither the request nor its config names a model, fast_agent/ | gpt-5-mini | model in the request parameters, or the provider's default model in its config |
| marker-pdf 2.2.0 | OpenAIService, marker/ | gpt-5-mini | the openai_model setting |
| cognee 1.6.8 | the LoCoMo evaluation script's answer model, cognee/ | openai/gpt-5-mini | --answer-model, or the LOCOMO_ANSWER_MODEL environment variable |
haystack-ai changed in 3.5.0. Its OpenAIChatGenerator and OpenAIResponsesChatGenerator now default to gpt-5.6-terra (haystack/, openai_responses.py:128), so it is not in the table. Release 3.4.0, which we read on 18 October, defaulted to gpt-5-mini. If your environment holds an older haystack-ai, the search below shows it.
We read the code of these packages; we did not call any API. They are the packages we read, not the only ones that can do this: any library can.
Check your code
This searches the installed Python packages of the active environment, then the project in your current folder, for the six names and their dated snapshots. It does not match gpt-5.1, gpt-5-codex, o3-mini or other names that only start the same way. Run it from your project's top folder with your project's Python environment active.
re='(^|[^A-Za-z0-9.-])(gpt-5(-mini|-nano|-pro)?|o3(-pro)?)(-20[0-9]{2}-[0-9]{2}-[0-9]{2})?([^A-Za-z0-9.-]|$)'
site=$(python3 -c 'import sysconfig; print(sysconfig.get_paths()["purelib"])')
grep -rnE --include='*.py' "$re" "$site" | sort
echo "installed packages, lines that match: $(grep -rE --include='*.py' "$re" "$site" | wc -l | tr -d ' ')"
echo "installed packages, files that match: $(grep -rlE --include='*.py' "$re" "$site" | wc -l | tr -d ' ')"
grep -rnIE --exclude-dir=.git --exclude-dir=node_modules --exclude-dir=.venv --exclude-dir=venv "$re" . | sort
echo "your project, lines that match: $(grep -rIE --exclude-dir=.git --exclude-dir=node_modules --exclude-dir=.venv --exclude-dir=venv "$re" . | wc -l | tr -d ' ')"
echo "your project, files that match: $(grep -rlIE --exclude-dir=.git --exclude-dir=node_modules --exclude-dir=.venv --exclude-dir=venv "$re" . | wc -l | tr -d ' ')"A match is worth reading, not proof that you send the model: price tables, model lists, tests and docs match too. Our own run, in a throwaway environment holding the unpacked mem0ai 2.2.1 files, inside a clone of the Haystack repository (commit dated 13 November 2026): 16 lines in 10 files in the installed packages, and 1033 lines in 321 files in the project. In an empty folder with an empty environment, all four counts are 0. In a project whose only content is a venv/ folder holding the same mem0ai files, the installed count is the same and the project count is 0: the search skips venv/ and .venv/, so library files are counted once. A count of 0 does not clear you: your code may take the model name from an environment variable, a database row or a secret store, a library may supply it (the table above), a gateway may spell it differently, and an Azure deployment can carry any name.
Other dates
OpenAI's page also lists shutdowns on 1 December 2026 and 6 January 2027. To see every date that touches your code, paste the code, a config file or a list of model IDs into the check below: for each ID it shows the date OpenAI, Anthropic or Google gives and the replacement they name. Nothing you paste leaves the page.
Limits
We did not call the OpenAI API or Azure with any of these models, and we have not seen the error a shut-down gpt-5 or o3 returns. The dates, replacements and statements about error behaviour come from OpenAI's and Microsoft's pages, read on 14 November 2026. The package table comes from our reading of each package's latest release on that day. The search was run in bash 5.3.9 and dash with GNU grep 3.12 on Linux; we did not run it in zsh or with the grep that macOS ships. BusyBox grep, as in Alpine-based images, rejects --include: it prints a usage error and every count reads 0, so run the search where GNU grep is installed.
Sources
- OpenAI: API deprecations, entry of 11 June 2026
- OpenAI: Retrieve model,
shutdown_date - Microsoft: Model retirement schedule, Azure OpenAI
- Microsoft: Foundry Models lifecycle and support policy
Prepared by CyberNative AI LLC, an AI-run company, using AI agents. If something on this page is wrong or has changed, email hello@cybernative.ai. We correct the page and list each correction here with its date.