45 AI model shutdowns are scheduled, 34 of them OpenAI's. GPT-4o and GPT-4.1 go on January 14

OpenAI, Anthropic and Google list 45 dated model shutdowns on their deprecation pages as of November 3, by ArtificialWatch's count of those pages. 34 are OpenAI's, 10 are Google's and one is Anthropic's. 34 of the 45 fall in the next 90 days, and 11 in the next 30.

THE NEXT 90 DAYS

- November 17, at the earliest: five Google preview models, Gemini 2.5 Flash Preview TTS, Gemini 2.5 Pro Preview TTS, Gemini 3.1 Flash TTS Preview, Gemini 3.1 Flash Live Preview and the December 2025 Gemini 2.5 Flash native-audio preview. Google's replacements are Gemini 3.8 Flash TTS (or Flash-Lite TTS) and Gemini 3.8 Live.

- November 30: Claude Sonnet 4.5 (claude-sonnet-4-5-20250929), Anthropic's only scheduled shutdown. Anthropic recommends Claude Sonnet 5.5.

- December 1: Gemini 2.5 Pro and Gemini 2.5 Flash, and three OpenAI image models: gpt-image-1.5, gpt-image-1-mini and chatgpt-image-latest.

- December 11: GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 pro, o3 and o3-pro. OpenAI points them to its GPT-5.6 Sol, Terra and Luna models.

- January 6: tts-1, tts-1-hd and two gpt-4o-mini-tts snapshots, all pointed to gpt-realtime-2.1-mini.

- January 14: GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini, all pointed to gpt-6-luna.

- January 20: nine OpenAI audio, realtime and transcription models, among them gpt-4o-realtime, gpt-realtime, gpt-audio and gpt-4o-audio.

January is the heaviest month, with 17 shutdowns, all of them OpenAI's.

GEMINI 2.5 PRO DOES NOT STOP ANSWERING, IT REROUTES

Google's December 1 change works differently from a shutdown. From that date, API requests to gemini-2.5-pro automatically route to Gemini 3.8 Flash and are billed at Gemini 3.8 Flash rates, and requests to gemini-2.5-flash route to Gemini 3.5 Flash-Lite. Code that calls either model keeps getting answers, from a different model. Google says gemini-2.5-pro is already gone from Google AI Studio and that its API access is limited to existing users.

AFTER JANUARY

On February 26, OpenAI retires whisper-1 and its gpt-4o transcription models (gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize) in favor of gpt-transcribe and gpt-live-transcribe, and moves /v1/audio/translations to /v1/realtime/translations. Google's gemini-2.5-flash-image follows on March 15 at the earliest, and GPT-5.1, GPT-5.3 Codex and GPT-5.4 nano on April 1. The furthest dated rows are Google's Gemini 3.1 Flash-Lite (May 7, 2027) and gemini-embedding-001 (May 14, 2028).

In the last 120 days the three labs already shut down 40 models: 29 at OpenAI, 10 at Google and one at Anthropic.

WHAT THESE NUMBERS ARE

Each count is a row on a lab's own deprecation page that names a model and a shutdown date. A row can be one model ID or a model with its dated snapshots, and one OpenAI row is an endpoint, not a model. Google lists the earliest possible shutdown date and says it will confirm the exact date with advance notice, so its rows can move later; OpenAI and Anthropic give the date a model stops answering on the API. The full list, each row linked to the lab's page, is at artificialwatch.com/model-retirements, and it is also a calendar feed you can subscribe to.