Check the OpenAI API Models Retiring from Late This Year into Early Next Year and Their Replacements
This page contains Coupang Partners affiliate links; the operator may earn a commission.
Key Takeaway
The first models to retire on the OpenAI API are those scheduled for October 23, 2026, including older GPT snapshots, o1, o3-mini, and o4-mini. The official page lists staged retirements through April 1, 2027.
What retires on which date?
I put the items developers use often from the official table in order of retirement date. Replacement model names are exactly as written in the table.
| Retirement date | Models affected | Recommended replacement |
|---|---|---|
| October 23, 2026 | Older GPT snapshots, fine-tuned models (ft-gpt-3.5-turbo, ft-gpt-4, etc.), o1, o3-mini, o4-mini, gpt-image-1 | Mostly gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna. For gpt-image-1, gpt-image-2.5-sunburst or flare |
| December 1, 2026 | gpt-image-1-mini, gpt-image-1.5, chatgpt-image-latest | gpt-image-2.5-sunburst or flare |
| December 11, 2026 | gpt-5-2025-08-07, gpt-5-mini, gpt-5-nano, o3-2025-04-16 | gpt-5.6-sol, terra, luna by tier |
| December 11, 2026 | gpt-5-pro-2025-10-06, o3-pro-2025-06-10 | gpt-5.6-sol with reasoning.mode set to pro |
| January 20, 2027 | Older realtime, audio, and transcription families | gpt-realtime-2.1, gpt-realtime-2.1-mini, gpt-audio-1.5 |
| February 26, 2027 | whisper-1, gpt-4o-transcribe family | gpt-live-transcribe or gpt-transcribe |
| April 1, 2027 | gpt-5.3-codex, gpt-5.1, gpt-5.4-nano | gpt-6-sol or gpt-6-luna |
What are the schedules for fine-tuning and voice?
The table says that from January 6, 2027, even existing active customers can no longer create new fine-tuning jobs. It says inference on fine-tuned models you have already built continues until the base model is retired. On the same day the tts-1, tts-1-hd, and gpt-4o-mini-tts snapshots also retire, and the recommended replacement is gpt-realtime-2.1-mini. Fine-tuned models are also included in the October 23 row, so check the base model and its retirement date separately.
What is the schedule for agent tools?
On November 30, the Evals platform, reusable prompts (the v1/prompts API), and Agent Builder retire. Evals becomes read-only from October 31. It tells you to move prompts into application code, and for Agent Builder the recommended path is the Agents SDK or ChatGPT Workspace Agents. ChatKit stays.
In what order should I check things now?
- Search your code, config files, and environment variables for model names and find any that appear in the table above.
- Starting with the item whose retirement date is closest, switch to the replacement model and compare the response format in a test environment.
- Cost and latency can differ, so check the replacement model's price separately in the official price table.
- Image and voice models can have different request parameters, so check the calling code too.
- Before the retirement date, make sure error alerts reach you, and make a rollback plan on the premise that you cannot go back to the old model name.
The official table only lists retiring models and recommended replacements, and does not guarantee that the replacement's output is the same. So compare results with your prompts and an evaluation set before switching. This article reflects the check date of 2026-10-08, and the table changes with new announcements, so right before you switch, check the date and replacement model for the relevant row again on the Deprecations page of the OpenAI API docs.
Frequently Asked Questions
What happens if I call an old model name after the retirement date?
The official page only lists retiring models and recommended replacements in a table, and the response format of calls after retirement is not confirmed there. Change the model name before the retirement date and test the responses.
Are the results the same if I switch to the replacement model?
The official table lists only the recommended replacement and does not guarantee identical output. It is safer to compare results with your prompts and an evaluation set before switching.
Can the replacement model names change?
Recommended replacements can be updated after the announcement. Check the official Deprecations page again right before you switch.
Sources & Verification Date
- Deprecations (OpenAI API docs) · Verified
All materials have been verified from official announcements and may have changed since verification.