Deprecated model
A deprecated model is one the provider still serves after announcing that support will end. It is not retired yet: requests keep working until the shutdown date, unless that provider's page says a narrower rule.
Last reviewed . Publisher: SunsetDiff. A source line names the day that page was read.
This dataset has 91 records with status deprecated. OpenAI records in that status: 63. Amazon Bedrock records in that status: 0. The full list is on the deprecated page.
OpenAI
OpenAI uses deprecation for the process of retiring a model or endpoint. When the announcement is made, the model becomes deprecated immediately. Every deprecated model or endpoint also has a shut down date. At shut down it is no longer accessible.
Sunset and shut down mean the same thing on that page. Legacy is a different word: no more updates, with deprecation expected later.
Source: OpenAI deprecations , checked October 22, 2026.
Anthropic
Anthropic's status table says Deprecated means the model is still functional but no longer recommended. Anthropic names a replacement and assigns a retirement date. Retired means it is no longer available and requests fail.
Those dates apply to the Claude API, Claude Platform on AWS, and Microsoft Foundry. Amazon Bedrock and Google Cloud publish their own schedules.
Source: Anthropic model deprecations , checked October 22, 2026.
Gemini API and Vertex AI
The Gemini API page says a deprecation is the announcement that Google no longer provides support and that the model will be shut down. After shutdown the model is turned off and the endpoint is no longer available. A date in the table is the earliest the model might be retired. The exact day is sent later, with advance notice.
The same page says some Gemini 2.5 models are not deprecated and stay up until further notice. Vertex lists a retirement date per model id. A listed Vertex date will not move earlier, though it may be extended.
Sources: Gemini API deprecations , Vertex model versions , checked October 22, 2026.
Amazon Bedrock
Bedrock's lifecycle pages do not use Deprecated as a state. The states are Active, Legacy, and End-of-Life (EOL). Legacy is the scheduled-retirement window. EOL is when requests fail.
SunsetDiff therefore does not file a Bedrock lifecycle row as status deprecated. Those rows are legacy until the EOL date has passed, then retired.
Sources: Bedrock model lifecycle , Bedrock legacy lifecycle , checked October 22, 2026.
Azure OpenAI
Azure's lifecycle table says Deprecated means existing customers can still create and manage deployments, and new customers cannot. An existing customer is an Azure subscription that has deployed that model version. A new subscription under the same tenant does not inherit access.
Legacy is an earlier optional stage: newer models exist, and you should plan to move. A model can skip Legacy and go from generally available to Deprecated. Retired is later: removed from service, and inference requests return 410 Gone.
Source: Foundry Models lifecycle , checked October 22, 2026.
Rows from this dataset
Dataset rows stored as deprecated. One per provider that has one, the soonest shutdown date first.
| Model | Provider | Stored status | Shutdown | Replacement |
|---|---|---|---|---|
| ft-babbage-002 | OpenAI | deprecated | 2026-10-23 | gpt-5.6-terra · diff |
| claude-sonnet-4-5-20250929 | Anthropic | deprecated | 2026-11-30 | claude-sonnet-5-5 · diff |
| gemini-2.5-flash-native-audio-preview-12-2025 | Google Gemini API | deprecated | 2026-11-17 | gemini-3.8-live · diff |
| gemini-3.6-flash | Google Vertex AI | deprecated | 2026-11-19 | gemini-3.8-flash · diff |
codex-mini (2025-05-16) codex-mini-2025-05-16 | Azure OpenAI | deprecated | 2026-11-15 | Not published |
Z.ai GLM 5.2 zai-glm-5-2 | Mistral | deprecated | 2026-10-31 | Not published |