OpenAI API Models End of Life (EOL) Dates and End of Support (EOS) Dates
OpenAI API Models end of life (EOL) and end of support (EOS) dates: 29 OpenAI API Models versions have reached EOL, including OpenAI API Models GPT-5.3 Chat (EOL date: Aug 10, 2026) — no longer receiving security updates or patches. OpenAI API Models chatgpt-image-latest reaches its End of Support date on Dec 01, 2026 — plan your upgrade path now. OpenAI API Models GPT-6.1 Terra is the current supported release — 35 versions in active support. Also see Azure Kubernetes Service end of life dates for related end-of-life information.
Currently Supported OpenAI API Models Releases - Active EOS/EOSL Support (35)
| Release | Aliases | Recommended Replacement | Released | Deprecated support | Active support |
|---|---|---|---|---|---|
| GPT-6.1 Terra | gpt-6.1-terra | - | (3 weeks and 5 days ago) | Active | Active |
| GPT-6.1 Sol | gpt-6.1-sol | - | (1 month and 3 weeks ago) | Active | Active |
| GPT-6 Luna | gpt-6-luna | N/A | (1 month and 4 weeks ago) | Active | Active |
| GPT-6 Sol | gpt-6-sol | N/A | (1 month and 4 weeks ago) | Active | Active |
| GPT-Image-2.5 Sunburst | gpt-image-2.5-sunburst, gpt-image-2.5-sunburst-2026-09-08 | N/A | (2 months ago) | Active | Active |
| GPT-Image-2.5 Flare | gpt-image-2.5-flare, gpt-image-2.5-flare-2026-09-08 | N/A | (2 months ago) | Active | Active |
| GPT-6 Astra | gpt-6-astra | N/A | (2 months ago) | Active | Active |
| GPT-5.6 Cyber | gpt-5.6-cyber | N/A | (3 months ago) | Active | Active |
| GPT-Transcribe | gpt-transcribe | N/A | (3 months ago) | Active | Active |
| GPT-Live-Transcribe | gpt-live-transcribe | N/A | (3 months ago) | Active | Active |
| GPT-5.6 Sol | gpt-5.6-sol | N/A | (4 months ago) | Active | Active |
| GPT-5.6 Terra | gpt-5.6-terra | N/A | (4 months ago) | Active | Active |
| GPT-5.6 Luna | gpt-5.6-luna | N/A | (4 months ago) | Active | Active |
| GPT-Live-1 | gpt-live-1, gpt-live-1 | N/A | (4 months ago) | Active | Active |
| GPT-Realtime-2.1 mini | gpt-realtime-2.1-mini | N/A | (4 months ago) | Active | Active |
| GPT-Realtime-2.1 | gpt-realtime-2.1 | N/A | (4 months ago) | Active | Active |
| GPT-Realtime-Whisper | gpt-realtime-whisper | N/A | (6 months ago) | Active | Active |
| GPT-Realtime-Translate | gpt-realtime-translate | N/A | (6 months ago) | Active | Active |
| GPT-Realtime-2 | gpt-realtime-2 | N/A | (6 months ago) | Active | Active |
| Chat Latest | chat-latest | N/A | (6 months ago) | Active | Active |
| GPT-5.5 Pro | gpt-5.5-pro-2026-04-23 | N/A | (6 months ago) | Active | Active |
| GPT-5.5 | gpt-5.5-2026-04-23 | N/A | (6 months ago) | Active | Active |
| GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | N/A | (8 months ago) | Active | Active |
| GPT-5.4 Pro | gpt-5.4-pro-2026-03-05 | N/A | (8 months ago) | Active | Active |
| GPT-5.4 | gpt-5.4-2026-03-05 | N/A | (8 months ago) | Active | Active |
| GPT-Realtime-1.5 | gpt-realtime-1.5 | N/A | (8 months ago) | Active | Active |
| GPT-Audio-1.5 | gpt-audio-1.5 | N/A | (8 months ago) | Active | Active |
| GPT-5.2 Pro | gpt-5.2-pro-2025-12-11 | N/A | (11 months ago) | Active | Active |
| GPT-5.2 | gpt-5.2-2025-12-11 | N/A | (11 months ago) | Active | Active |
| gpt-oss-20b | gpt-oss-20b | N/A | (1 year and 3 months ago) | Active | Active |
| gpt-oss-120b | gpt-oss-120b | N/A | (1 year and 3 months ago) | Active | Active |
| omni-moderation | omni-moderation-2024-09-26 | N/A | (2 years and 1 month ago) | Active | Active |
| text-embedding-3-small | text-embedding-3-small | N/A | (2 years and 9 months ago) | Active | Active |
| text-embedding-3-large | text-embedding-3-large | N/A | (2 years and 9 months ago) | Active | Active |
| text-embedding-ada-002 | text-embedding-ada-002 | N/A | (4 years and 9 months ago) | Active | Active |
OpenAI API Models Releases Nearing End of Life - Support Ending Soon (28)
| Release | Aliases | Recommended Replacement | Released | Deprecated support | Active support |
|---|---|---|---|---|---|
| chatgpt-image-latest | chatgpt-image-latest | gpt-image-2.6-sunburst or gpt-image-2.6-flare | (1 year and 7 months ago) | (in 1 week and 3 days) | (5 months ago) |
| GPT Image 1 mini | gpt-image-1-mini | gpt-image-2.6-sunburst or gpt-image-2.6-flare | (1 year and 1 month ago) | (in 1 week and 3 days) | (5 months ago) |
| GPT Image 1.5 | gpt-image-1.5-2025-12-16 | gpt-image-2.6-sunburst or gpt-image-2.6-flare | (11 months ago) | (in 1 week and 3 days) | (5 months ago) |
| o3-pro | o3-pro-2025-06-10 | gpt-6.1-sol ( reasoning.mode: pro ) in the Responses API | (1 year and 7 months ago) | (in 2 weeks and 6 days) | (5 months ago) |
| o3 | o3-2025-04-16 | gpt-6.1-sol | (1 year and 7 months ago) | (in 2 weeks and 6 days) | (5 months ago) |
| GPT-5 pro | gpt-5-pro-2025-10-06 | gpt-6.1-sol ( reasoning.mode: pro ) in the Responses API | (1 year and 3 months ago) | (in 2 weeks and 6 days) | (5 months ago) |
| GPT-5 nano | gpt-5-nano-2025-08-07 | gpt-6-luna | (1 year and 3 months ago) | (in 2 weeks and 6 days) | (5 months ago) |
| GPT-5 mini | gpt-5-mini-2025-08-07 | gpt-6.1-terra | (1 year and 3 months ago) | (in 2 weeks and 6 days) | (5 months ago) |
| GPT-5 | gpt-5-2025-08-07 | gpt-6.1-sol | (1 year and 3 months ago) | (in 2 weeks and 6 days) | (5 months ago) |
| TTS-1 HD | tts-1-hd | gpt-realtime-2.1-mini | (3 years ago) | (in 1 month and 2 weeks) | (1 month and 2 weeks ago) |
| TTS-1 | tts-1 | gpt-realtime-2.1-mini | (3 years ago) | (in 1 month and 2 weeks) | (1 month and 2 weeks ago) |
| GPT-4o mini TTS | gpt-4o-mini-tts-2025-03-20, gpt-4o-mini-tts-2025-12-15 | gpt-realtime-2.1-mini | (1 year and 8 months ago) | (in 1 month and 2 weeks) | (1 month and 2 weeks ago) |
| GPT-4o | gpt-4o-2024-11-20, gpt-4o-2024-08-06, gpt-4o-2024-05-13 | gpt-6-luna | (2 years and 6 months ago) | (in 1 month and 3 weeks) | (1 month and 1 week ago) |
| GPT-4o mini | gpt-4o-mini-2024-07-18 | gpt-6-luna | (2 years and 4 months ago) | (in 1 month and 3 weeks) | (1 month and 1 week ago) |
| GPT-4.1 mini | gpt-4.1-mini-2025-04-14 | gpt-6-luna | (1 year and 7 months ago) | (in 1 month and 3 weeks) | (1 month and 1 week ago) |
| GPT-4.1 | gpt-4.1-2025-04-14 | gpt-6-luna | (1 year and 7 months ago) | (in 1 month and 3 weeks) | (1 month and 1 week ago) |
| GPT-4o Mini Transcribe | gpt-4o-mini-transcribe-2025-03-20, gpt-4o-mini-transcribe-2025-12-15 | gpt-4o-mini-transcribe-2025-12-15 | (1 year and 8 months ago) | (in 1 month and 4 weeks) | (4 months ago) |
| GPT-Realtime | gpt-realtime-2025-08-28 | gpt-realtime-2.1 | (1 year and 2 months ago) | (in 1 month and 4 weeks) | (4 months ago) |
| GPT-Realtime mini | gpt-realtime-mini-2025-10-06, gpt-realtime-mini-2025-12-15 | gpt-realtime-2.1-mini | (1 year and 1 month ago) | (in 1 month and 4 weeks) | (4 months ago) |
| GPT-Audio mini | gpt-audio-mini-2025-10-06, gpt-audio-mini-2025-12-15 | gpt-audio-1.5 | (10 months ago) | (in 1 month and 4 weeks) | (4 months ago) |
| GPT-Audio | gpt-audio-2025-08-28 | gpt-audio-1.5 | (10 months ago) | (in 1 month and 4 weeks) | (4 months ago) |
| GPT-Image-2 | gpt-image-2-2026-04-21 | gpt-image-2.6-sunburst or gpt-image-2.6-flare | (7 months ago) | (in 2 months) | (1 week and 4 days ago) |
| Whisper | whisper-1 | gpt-live-transcribe or gpt-transcribe | (4 years and 2 months ago) | (in 3 months) | (2 months ago) |
| GPT-4o Transcribe | gpt-4o-transcribe | gpt-live-transcribe or gpt-transcribe | (1 year and 8 months ago) | (in 3 months) | (2 months ago) |
| GPT-4o Transcribe Diarize | gpt-4o-transcribe-diarize | gpt-live-transcribe or gpt-transcribe | (1 year and 1 month ago) | (in 3 months) | (2 months ago) |
| GPT-5.1 | gpt-5.1-2025-11-13 | gpt-6-sol | (1 year ago) | (in 4 months) | (1 month and 2 weeks ago) |
| GPT-5.3-Codex | gpt-5.3-codex | gpt-6-sol | (9 months ago) | (in 4 months) | (1 month and 2 weeks ago) |
| GPT-5.4 nano | gpt-5.4-nano-2026-03-17 | gpt-6-luna | (8 months ago) | (in 4 months) | (1 month and 2 weeks ago) |
OpenAI API Models End of Life (EOL) Releases - No Longer Supported (29)
| Release | Aliases | Recommended Replacement | Released | Deprecated support | Active support |
|---|---|---|---|---|---|
| GPT-5.3 Chat | gpt-5.3-chat-latest | gpt-5.6-sol | (8 months ago) | (3 months ago) | (6 months ago) |
| GPT-5.2-Codex | gpt-5.2-codex | gpt-5.6-sol | (11 months ago) | (3 months ago) | (6 months ago) |
| GPT-5.2 Chat | gpt-5.2-chat-latest | gpt-5.6-sol | (11 months ago) | (3 months ago) | (6 months ago) |
| GPT-5.1-Codex-Max | gpt-5.1-codex-max | gpt-5.6-sol | (1 year ago) | (3 months ago) | (6 months ago) |
| GPT-5.1-Codex-Mini | gpt-5.1-codex-mini | gpt-5.6-terra | (1 year ago) | (3 months ago) | (6 months ago) |
| GPT-5.1-Codex | gpt-5.1-codex | gpt-5.6-sol | (1 year ago) | (3 months ago) | (6 months ago) |
| GPT-5.1 Chat | gpt-5.1-chat-latest | gpt-5.6-sol | (1 year ago) | (3 months ago) | (6 months ago) |
| Sora 2 Pro | sora-2-pro-2025-10-06, sora-2-pro | — | (1 year and 1 month ago) | (1 month and 4 weeks ago) | (7 months ago) |
| Sora 2 | sora-2-2025-12-08, sora-2-2025-10-06, sora-2 | — | (1 year and 1 month ago) | (1 month and 4 weeks ago) | (7 months ago) |
| GPT-5-Codex | gpt-5-codex | gpt-5.6-sol | (1 year and 2 months ago) | (3 months ago) | (6 months ago) |
| GPT-5 Chat | gpt-5-chat-latest | gpt-5.6-sol | (1 year and 3 months ago) | (3 months ago) | (6 months ago) |
| codex-mini-latest | codex-mini-latest | gpt-5-codex-mini | (1 year and 6 months ago) | (9 months ago) | (1 year ago) |
| o4-mini | o4-mini-2025-04-16 | gpt-5.6-terra | (1 year and 7 months ago) | (4 weeks and 1 day ago) | (6 months ago) |
| GPT-4.1 nano | gpt-4.1-nano-2025-04-14 | gpt-5.6-luna | (1 year and 7 months ago) | (4 weeks and 1 day ago) | (6 months ago) |
| GPT Image 1 | gpt-image-1 | gpt-image-2.5-sunburst or gpt-image-2.5-flare | (1 year and 7 months ago) | (4 weeks and 1 day ago) | (6 months ago) |
| o1-pro | o1-pro-2025-03-19 | gpt-5.6-sol ( reasoning.mode: pro ) | (1 year and 8 months ago) | (4 weeks and 1 day ago) | (6 months ago) |
| o4-mini-deep-research | o4-mini-deep-research-2025-06-26 | gpt-5.6-sol | (1 year and 9 months ago) | (3 months ago) | (6 months ago) |
| o3-deep-research | o3-deep-research-2025-06-26 | gpt-5.6-sol | (1 year and 9 months ago) | (3 months ago) | (6 months ago) |
| o3-mini | o3-mini-2025-01-31 | gpt-5.6-sol | (1 year and 9 months ago) | (4 weeks and 1 day ago) | (6 months ago) |
| o1 | o1-2024-12-17 | gpt-5.6-sol | (1 year and 11 months ago) | (4 weeks and 1 day ago) | (6 months ago) |
| o1-mini | o1-mini-2024-09-12 | o4-mini | (2 years and 2 months ago) | (1 year ago) | (1 year and 6 months ago) |
| ChatGPT-4o | chatgpt-4o-latest | gpt-5.1-chat-latest | (2 years and 3 months ago) | (9 months ago) | (1 year ago) |
| GPT-4 Turbo | gpt-4-turbo-2024-04-09 | gpt-5.6-sol | (3 years ago) | (4 weeks and 1 day ago) | (6 months ago) |
| davinci-002 | davinci-002 | gpt-5.6-terra | (3 years and 2 months ago) | (1 month and 3 weeks ago) | (1 year and 1 month ago) |
| babbage-002 | babbage-002 | gpt-5.6-terra | (3 years and 2 months ago) | (1 month and 3 weeks ago) | (1 year and 1 month ago) |
| GPT-4 | gpt-4-0613, gpt-4-0314 | gpt-5.6-sol | (3 years and 8 months ago) | (4 weeks and 1 day ago) | (6 months ago) |
| GPT-3.5 Turbo | gpt-3.5-turbo-0125, gpt-3.5-turbo-1106, gpt-3.5-turbo-instruct | gpt-5.6-terra | (3 years and 8 months ago) | (4 weeks and 1 day ago) | (6 months ago) |
| text-moderation-stable | text-moderation-007 | omni-moderation | (4 years and 3 months ago) | (1 year ago) | (1 year and 6 months ago) |
| text-moderation | text-moderation-007 | omni-moderation | (4 years and 3 months ago) | (1 year ago) | (1 year and 6 months ago) |
Every OpenAI API models end of life date is published in advance, and missing one means your requests start failing. The tables below pair each deprecated model with its shutdown date and the replacement OpenAI recommends, so developers can plan a migration well before the cutoff. They cover upcoming and past retirements, Azure OpenAI dates, and a checklist for your production API workloads, all verified against OpenAI's documentation on October 27, 2026.
OpenAI API Models End of Life: How the Model Lifecycle Works
As OpenAI launches safer and more capable models, it regularly retires older ones, so software that relies on a specific model version may need occasional updates. A deprecation is the announcement, and every model reaches EOL sooner or later; the shutdown is the day the endpoint stops answering. In between, the model keeps working while you test a replacement.
Deprecation Notice Periods for Each Model Type
The notice period depends on the kind of model, unless safety or compliance concerns force a faster timeline. In some cases a customer can arrange dedicated capacity for continued access after a shutdown by contacting OpenAI's sales team.
- Generally available models: at least 6 months of notice.
- Specialized variants of generally available models, such as chat variants and deep research variants like
o3-deep-research: at least 3 months. - Preview models: as little as 2 weeks, for example
computer-use-preview. Skip them for business-critical production workloads unless you can migrate on short notice.
Where OpenAI Announces Retirements
When OpenAI sends a deprecation notice, it emails customers who are actively using the model and records the change in its documentation. Larger changes also get blog posts. Check that email address on your account, because the email is the only direct warning most teams receive.
Deprecation vs. Legacy and Shutdown: What Each Term Means
A model or endpoint is deprecated the moment OpenAI announces its removal, and every deprecated model gets a shutdown date. "Sunset" and "shut down" mean the same thing: the model is no longer accessible. A legacy model only stops receiving updates, and you should expect it to be deprecated eventually. Trackers such as endoflife.date show this as two columns, active support and deprecated support, for each end-of-life (EOL) entry.
Checking Each Model's Shutdown Date Through the API
To monitor retirements programmatically, call the List models endpoint (GET /v1/models) and read each model's optional shutdown_date field. It holds a YYYY-MM-DD string, or null when no shutdown date has been announced. The Retrieve a model endpoint returns the same field, and a scheduled job that alerts on it gives you an early timeline without reading any announcement.
Upcoming Deprecations: OpenAI Models and Their Sunset Dates
This calendar of OpenAI deprecations lists models that are deprecated today and scheduled for removal, newest announcements first, so every upcoming cutoff is visible at a glance. Each row shows the shutdown date and the recommended replacement. Fine-tuned versions follow the same schedule as their base models, and the same deprecation applies to every dated version of a model name.
GPT-4o, GPT-4.1 and GPT-5 Series Retirements
On October 14, 2026, OpenAI notified developers that four GPT-4 series models leave the API on January 14, 2027.
| Model | Shutdown date | Recommended replacement | Lifecycle status |
|---|---|---|---|
GPT-4o (gpt-4o-2024-08-06, gpt-4o-2024-11-20) | Jan 14, 2027 | gpt-6-luna | Deprecated |
GPT-4o mini (gpt-4o-mini-2024-07-18) | Jan 14, 2027 | gpt-6-luna | Deprecated |
GPT-4.1 (gpt-4.1-2025-04-14) | Jan 14, 2027 | gpt-6-luna | Deprecated |
GPT-4.1 mini (gpt-4.1-mini-2025-04-14) | Jan 14, 2027 | gpt-6-luna | Deprecated |
Older GPT-5 and o3 snapshots go next. The June 11, 2026 notice set a December 11, 2026 cutoff for these models:
| Model | Shutdown date | Recommended replacement | Lifecycle status |
|---|---|---|---|
| gpt-5-2025-08-07 | Dec 11, 2026 | gpt-5.6-sol | Deprecated |
| gpt-5-mini-2025-08-07 | Dec 11, 2026 | gpt-5.6-terra | Deprecated |
| gpt-5-nano-2025-08-07 | Dec 11, 2026 | gpt-5.6-luna | Deprecated |
| gpt-5-pro-2025-10-06 | Dec 11, 2026 | gpt-5.6-sol (reasoning.mode: pro) | Deprecated |
| o3-2025-04-16 | Dec 11, 2026 | gpt-5.6-sol | Deprecated |
| o3-pro-2025-06-10 | Dec 11, 2026 | gpt-5.6-sol (reasoning.mode: pro) | Deprecated |
The October 1, 2026 notice gave six months of warning for three more models, all leaving on April 1, 2027:
| Model | Shutdown date | Recommended replacement | Lifecycle status |
|---|---|---|---|
| gpt-5.3-codex | Apr 1, 2027 | gpt-6-sol | Deprecated |
| gpt-5.4-nano-2026-03-17 | Apr 1, 2027 | gpt-6-luna | Deprecated |
| gpt-5.1-2025-11-13 | Apr 1, 2027 | gpt-6-sol | Deprecated |
Realtime, Text-to-Speech and Transcription Models
A July 20, 2026 notice retires the legacy audio and realtime families on January 20, 2027.
| Model family | Shutdown date | Recommended replacement | Lifecycle status |
|---|---|---|---|
| gpt-realtime | Jan 20, 2027 | gpt-realtime-2.1 | Deprecated |
| gpt-4o-realtime | Jan 20, 2027 | gpt-realtime-2.1 | Deprecated |
| gpt-realtime-mini | Jan 20, 2027 | gpt-realtime-2.1-mini | Deprecated |
| gpt-4o-mini-realtime | Jan 20, 2027 | gpt-realtime-2.1-mini | Deprecated |
| gpt-audio | Jan 20, 2027 | gpt-audio-1.5 | Deprecated |
| gpt-4o-audio | Jan 20, 2027 | gpt-audio-1.5 | Deprecated |
| gpt-audio-mini | Jan 20, 2027 | gpt-audio-1.5 | Deprecated |
| gpt-4o-mini-audio | Jan 20, 2027 | gpt-audio-1.5 | Deprecated |
The text-to-speech family follows on January 6, 2027, and the transcription models, including whisper-1, on February 26, 2027. The /v1/audio/translations endpoint moves to v1/realtime/translations on the same day.
| Model | Shutdown date | Recommended replacement | Lifecycle status |
|---|---|---|---|
| tts-1 | Jan 6, 2027 | gpt-realtime-2.1-mini | Deprecated |
| tts-1-hd | Jan 6, 2027 | gpt-realtime-2.1-mini | Deprecated |
| gpt-4o-mini-tts | Jan 6, 2027 | gpt-realtime-2.1-mini | Deprecated |
| whisper-1 | Feb 26, 2027 | gpt-live-transcribe or gpt-transcribe | Deprecated |
| gpt-4o-transcribe | Feb 26, 2027 | gpt-live-transcribe or gpt-transcribe | Deprecated |
| gpt-4o-mini-transcribe | Feb 26, 2027 | gpt-live-transcribe or gpt-transcribe | Deprecated |
| gpt-4o-transcribe-diarize | Feb 26, 2027 | gpt-live-transcribe or gpt-transcribe | Deprecated |
GPT Image Models Leaving the API
A June 2, 2026 notice removes the older GPT Image models on December 1, 2026, including the chatgpt-image-latest alias.
| Model | Shutdown date | Recommended replacement | Lifecycle status |
|---|---|---|---|
| gpt-image-1-mini | Dec 1, 2026 | gpt-image-2.5-sunburst or gpt-image-2.5-flare | Deprecated |
| gpt-image-1.5 | Dec 1, 2026 | gpt-image-2.5-sunburst or gpt-image-2.5-flare | Deprecated |
| chatgpt-image-latest | Dec 1, 2026 | gpt-image-2.5-sunburst or gpt-image-2.5-flare | Deprecated |
Evals, Agent Builder and Prompt Platform Shutdowns
Some retirements hit products rather than models. Existing evals become read-only on October 31, 2026, and the three platform features below shut down on November 30, 2026. Self-serve fine-tuning is also winding down: active customers can no longer create new fine-tuning jobs after January 6, 2027, and inference on a fine-tuned model ends only when its base model is deprecated, so check every application that depends on one.
| Product | Shutdown date | Recommended replacement | Lifecycle status |
|---|---|---|---|
Reusable prompts (v1/prompts API) | Nov 30, 2026 | Prompt text in your application code | Deprecated |
| Evals platform | Nov 30, 2026 | Promptfoo | Deprecated |
| Agent Builder | Nov 30, 2026 | Agents SDK or ChatGPT Workspace Agents | Deprecated |
Past Deprecations: Retired Models and Legacy Snapshots
These models are already gone, or their deprecated support has ended. Keeping the record matters because old model ids still turn up in codebases and old tutorials.
Legacy GPT Snapshots Shut Down in October 2026
The legacy snapshots OpenAI announced on April 22, 2026 stopped working on October 23, 2026. Calls to gpt-3.5-turbo and its dated snapshot now fail, and earlier cutoffs hit the Codex family and other snapshots in July and August.
| Model snapshot | Shutdown date | Recommended replacement | Lifecycle status |
|---|---|---|---|
gpt-3.5-turbo-0125 (gpt-3.5-turbo, gpt-3.5-turbo-completions) | Oct 23, 2026 | gpt-5.6-terra | Retired |
| gpt-4-0613 and gpt-4-turbo | Oct 23, 2026 | gpt-5.6-sol | Retired |
| gpt-4.1-nano | Oct 23, 2026 | gpt-5.6-luna | Retired |
| o4-mini | Oct 23, 2026 | gpt-5.6-terra | Retired |
| o1-pro | Oct 23, 2026 | gpt-5.6-sol (reasoning.mode: pro) | Retired |
| gpt-image-1 | Oct 23, 2026 | gpt-image-2.5-sunburst or gpt-image-2.5-flare | Retired |
| gpt-5.4-cyber | Oct 1, 2026 | The most capable cyber model available to you | Retired |
| gpt-5.2-chat-latest, gpt-5.3-chat-latest | Aug 10, 2026 | gpt-5.6-sol | Retired |
| gpt-5-codex, gpt-5.2-codex | Jul 23, 2026 | gpt-5.6-sol | Retired |
| gpt-5.1-codex, gpt-5.1-codex-max | Jul 23, 2026 | gpt-5.6-sol | Retired |
| gpt-5.1-codex-mini | Jul 23, 2026 | gpt-5.6-terra | Retired |
| o4-mini-deep-research, o3-deep-research | Jul 23, 2026 | gpt-5.6-sol | Retired |
Fine-Tuned Models After Base Model Retirement
Fine-tuned models retire in two phases, training and deployment. OpenAI's self-serve platform stopped accepting first-time fine-tuning customers on May 7, 2026, and on July 2, 2026 it also closed job creation to organizations with no inference on a fine-tuned model in the previous 60 days. Because inference ends only at base model retirement, a fine-tuned GPT-4o keeps running until January 14, 2027, then needs retraining on a supported base.
Embedding and Preview Models
OpenAI's deprecations page also records first-generation InstructGPT, base GPT and text embedding models among past deprecations. Current embedding models are the quiet exception: on Azure they stay generally available well past the other dates.
| Model | Type | Lifecycle status | Azure retirement date |
|---|---|---|---|
| text-embedding-3-large | Embedding model | GA | Feb 9, 2028 |
| text-embedding-3-small | Embedding model | GA | Feb 9, 2028 |
| text-embedding-ada-002 | Embedding model | GA | Feb 9, 2028 |
Preview models go first and fastest. On Azure, preview sora-2 retires October 15, 2026, and the original gpt-image-1 preview retires October 23, 2026. Open-weight models such as gpt-oss-120b and gpt-oss-20b show no announced retirement, since you host them yourself.
Azure OpenAI Model Retirement Schedule and Other Vendors
The same OpenAI model can carry different dates on different platforms. This is the single most important detail for anyone calling models through a cloud vendor: each platform follows its own schedule, and its dates win for requests you send there.
Azure OpenAI and Microsoft Foundry Retirement Dates
Microsoft Foundry publishes a retirement schedule for Azure OpenAI with a lifecycle stage, a retirement date and a replacement column that also names the target for automatic upgrades. A few entries land earlier than OpenAI's own dates, and a retirement date there is firm for any deployment you run on that platform. Read the schedule alongside the deprecation notices OpenAI sends for its own API, because the two lists can disagree for the same model version.
| Model | Version | Lifecycle | Retirement date | Replacement |
|---|---|---|---|---|
| codex-mini | 2025-05-16 | Deprecated | 2026-11-15 | None listed |
| o1, o3, o3-pro | 2024-12-17, 2025-04-16 | Deprecated | 2026-11-19 | gpt-5.6-sol |
| o3-mini, o4-mini | 2025-01-31, 2025-04-16 | Deprecated | 2026-11-19 | gpt-5.6-terra |
| gpt-4o | 2024-05-13 | Deprecated | 2026-12-09 | gpt-5.6-sol |
| gpt-4o | 2024-11-20 | Legacy | 2027-04-14 | gpt-5.1 |
| gpt-image-1.5 | 2025-12-16 | GA | 2026-12-16 | None listed |
Azure's lifecycle stages run from preview and GA through legacy and deprecated to retired, and region does not change them.
Anthropic, Google and Amazon Bedrock Lifecycle Policies
If your stack mixes providers, expect different vocabulary for the same two ideas: still callable but scheduled to end, and no longer callable.
| Vendor | Still callable | No longer callable |
|---|---|---|
| OpenAI | Deprecated, legacy | Shutdown, sunset |
| Anthropic | Deprecated, legacy | Retired |
| Google Gemini API | Scheduled with a shutdown date | Shutdown |
| Google Vertex AI | Legacy stable | Retired |
| Amazon Bedrock | Legacy | EOL |
On Amazon Bedrock, a legacy model sits in that state for at least six months before its EOL date, and new customers cannot start using it. Models with later EOL dates may also move to higher provider-set pricing during an extended access period. Anthropic marks a model deprecated while it still works, assigns a retirement date and a recommended replacement, and calls it retired once requests fail. Anthropic models are also offered through other clouds, so an Anthropic model can have different dates depending on where you call it. Google runs two systems: the Gemini API has no separate deprecated state, only a shutdown date, while Google Vertex AI separates the latest stable version from a legacy stable one. A Google model can therefore carry one date in the Gemini API and another on Google Cloud. On Amazon Bedrock, EOL means the model is unavailable in every AWS Region, and Amazon Bedrock shows the Legacy date on each model card. Each vendor documents its own schedule:
- Anthropic keeps a model deprecations page for the Anthropic API.
- Google splits Google AI for Developers and Google Cloud lifecycle pages.
- Amazon Bedrock lists a model lifecycle page in its user guide.
Planning a Migration Before the Shutdown Date
A date only becomes actionable once you know which model identifiers your systems call. Treat OpenAI API models end of life as a recurring operations task, not a one-off scramble: add each cutoff from the calendar above to your team calendar, and migrate in order of the earliest shutdown first. The OpenAI API, the Gemini API and the Anthropic API all follow this same pattern, so one process covers every provider.
Choosing a Recommended Replacement
The recommended replacement is OpenAI's suggestion, not a promise of identical behavior. The listings point most mini and nano models to the luna or terra tiers and most pro and Codex models to sol. Pin an exact version string rather than a floating alias where you can, since each version has its own dates. Compare quality, latency and cost on your own prompts, and remember you need not wait for a replacement to be listed before you evaluate another compatible model.
Migration Checklist for Developers
- Developers should inventory every model id in code, configuration and logs, including fine-tuned models and any version pinned in an API request.
- Match each id to a shutdown date in the tables above.
- Run your own evaluation prompts through the recommended replacement and compare the results.
- Ship the switch several weeks before the cutoff and watch error rates after the change; a version that works in testing can still behave differently under production traffic.