Book a scoping call
Menu
Model retirements

Which of your models shut down, when, and what breaks

Shutdown dates from OpenAI, Anthropic, Google, Azure AI Foundry and AWS Bedrock, with the replacement and the code change each one needs. Last checked against the providers' own pages on October 25, 2026.

Check your exposure

Paste code, an .env file, a gateway config or grep output. The check runs in your browser: nothing you paste is sent anywhere.

Ctrl or ⌘ + Enter to run

The calendar

Sorted by date. Each row links to the provider's own page. Dates in grey have passed; if your code still calls those models, it is failing now.

Date Provider What retires Replacement What changes in your code
Aug 26, 2026 Passed OpenAI Assistants API removed. Responses API and Conversations API Calls to beta.assistants and beta.threads fail. Assistant, thread and run objects have to be rewritten against the Responses API, with conversation state in the Conversations API. Source Fix
Oct 20, 2026 Passed Google Vertex AI gemini-2.5-pro, gemini-2.5-flash and gemini-2.5-flash-lite retired on Vertex AI. gemini-3.8-flash, gemini-3.5-flash or gemini-3.5-flash-lite Pinned 2.5 model names stop resolving on Vertex AI. Re-test prompts on the replacement before switching. Source Fix
Oct 21, 2026 Passed Anthropic API version 2026-10-21 removes temperature, top_p and top_k. The new SDK major versions (Python 2.0.0, TypeScript 1.0.0) default to it. Remove sampling parameters from requests Any request that sets temperature, top_p or top_k returns a 400 error once you upgrade the SDK or move to the 5.x models. Source Fix
Oct 23, 2026 Passed OpenAI Shut down: gpt-3.5-turbo, gpt-4, gpt-4-1106-preview, gpt-4-turbo, gpt-4.1-nano, gpt-4o-2024-05-13, gpt-image-1, o1, o1-pro, o3-mini, o4-mini, babbage-002, davinci-002, and fine-tunes of these. gpt-5.6-sol, gpt-5.6-terra or gpt-5.6-luna; gpt-image-2.5-sunburst or gpt-image-2.5-flare for images Calls return 404 model_not_found. Several teams fixed this by moving to gpt-4o or gpt-image-1-mini, which shut down on January 14 and December 1. Source Fix
Oct 31, 2026 Passed OpenAI Existing Evals become read-only. Promptfoo (OpenAI publishes a migration guide) No new eval runs on the platform. Export eval data and graders before the November 30 shutdown. Source Fix
Nov 1, 2026 Passed Microsoft Power Platform AI Builder capacity add-ons and seeded AI Builder credits reach end of life. Copilot Credits No code change, but AI Builder actions in Power Apps and Power Automate start drawing on Copilot Credits: a cost change. Source Fix
Nov 13, 2026 Passed AWS Bedrock Claude 3 Haiku, Claude 3 Sonnet, Claude 3.5 Sonnet (v1 and v2) and Claude 3.7 Sonnet reach end of life in non-US regions only (ap-*, eu-west-2, ca-central-1, sa-east-1, eu-central-2). Newer Claude models US companies serving users from those regions lose these models. US regions are not affected on this date. Source Fix
Nov 15, 2026 Passed Azure AI Foundry codex-mini (2025-05-16) and claude-haiku-4-5 retire on Azure. See the Azure retirement schedule Azure deployments of these models stop. On Anthropic's own API, claude-haiku-4-5 is still active. Source Fix
Nov 17, 2026 Passed Google Gemini API Preview TTS models shut down: gemini-3.1-flash-tts-preview, gemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts. gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts Voice and text-to-speech features stop. Check voice similarity on the replacement before switching. Source Fix
Nov 17, 2026 Passed Google Gemini API Live preview models shut down: gemini-3.1-flash-live-preview and gemini-2.5-flash-native-audio-preview-12-2025. gemini-3.8-live Live and voice-agent sessions on these models stop connecting. Source Fix
Nov 19, 2026 Passed Google Vertex AI gemini-3.6-flash retires on Vertex AI. gemini-3.8-flash Pinned calls stop resolving. Re-test prompts on the replacement. Source Fix
Nov 19, 2026 Passed Azure AI Foundry o1, o1-pro, o3, o3-deep-research, o3-mini, o3-pro and o4-mini retire on Azure, as does Llama-4-Maverick-17B-128E-Instruct-FP8. gpt-5.6-sol or gpt-5.6-terra Azure OpenAI reasoning deployments stop. This is earlier than OpenAI’s own December 11 date for o3 and o3-pro. Source Fix
Nov 24, 2026 Passed Azure AI Foundry claude-opus-4-5 retires on Azure. On Anthropic's own API it is still active, 'not sooner than Nov 24, 2026'. See the Azure retirement schedule Azure Claude deployments stop. Anthropic API users should expect a formal notice. Source Fix
Nov 26, 2026 Passed AWS Bedrock AI21 Jamba 1.5 Large and Jamba 1.5 Mini reach end of life. See the Bedrock model lifecycle page Bedrock calls to these model IDs fail. Source Fix
Nov 30, 2026 Passed OpenAI Evals dashboard and API shut down. Promptfoo, or your own harness in CI Eval runs, graders and history on the platform are gone. Anything not exported is lost. Source Fix
Nov 30, 2026 Passed OpenAI Agent Builder shuts down. ChatKit remains. Agents SDK or ChatGPT Workspace Agents No-code agents built in Agent Builder have to be rebuilt in code or in Workspace Agents. Source Fix
Nov 30, 2026 Passed OpenAI The v1/prompts API and reusable prompt objects shut down. Prompt content kept in application code Requests that reference a stored prompt ID (pmpt_…) fail. Prompt text and variables have to move into your code. Source Fix
Nov 30, 2026 Passed Anthropic claude-sonnet-4-5-20250929 retires. claude-sonnet-5-5 Calls to Sonnet 4.5 fail. Moving to a 5.x model also means temperature, top_p and top_k are rejected, so remove them in the same change. Source Fix
Nov 30, 2026 Passed Azure AI Foundry claude-sonnet-4-5 retires on Azure. claude-sonnet-5-5 Azure Claude deployments of Sonnet 4.5 stop. Source Fix
Nov 30, 2026 Passed AWS Bedrock TwelveLabs Marengo Embed v2.7 reaches end of life. See the Bedrock model lifecycle page Video-embedding calls fail. Stored embeddings from v2.7 are not comparable with a new model’s; plan a re-index. Source Fix
Dec 1, 2026 Passed OpenAI Shut down: gpt-image-1-mini, gpt-image-1.5 (including gpt-image-1.5-2025-12-16) and chatgpt-image-latest. gpt-image-2.5-sunburst or gpt-image-2.5-flare Image generation and editing calls fail, including in apps that only just moved off gpt-image-1. Source Fix
Dec 9, 2026 Passed Azure AI Foundry gpt-4o version 2024-05-13 auto-upgrades to gpt-5.6-sol on Standard deployments. Provisioned deployments are not upgraded. Ministral-3B moves to ministral-4b and mistral-medium-2505 to mistral-medium-3-5. gpt-5.6-sol; ministral-4b; mistral-medium-3-5 Standard deployments change behavior without a code change. Provisioned deployments break. Source Fix
Dec 11, 2026 Passed OpenAI Shut down: gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3 and o3-pro (aliases and snapshots). gpt-5.6-sol, gpt-5.6-terra or gpt-5.6-luna; gpt-5.6-sol with reasoning.mode: pro for the pro models Calls fail. Moving the pro models means setting the reasoning mode on the replacement rather than changing a name. Source Fix
Dec 11, 2026 Passed ChatGPT Custom GPTs retire. Enterprise workspaces with an approved deferral have until February 11, 2027. Plugins ("Migrate to plugin") Instructions become a skill and knowledge files become reference files, but custom actions, the selected model and sharing settings do not transfer. Source Fix
Dec 13, 2026 Passed Google Vertex AI gemini-live-2.5-flash-native-audio retires on Vertex AI. gemini-3.8-live Vertex voice agents on this model stop connecting. Source Fix
Dec 15, 2026 Passed Azure AI Foundry tts, tts-hd and whisper retire on Azure. See the Azure retirement schedule Azure speech deployments stop. Plan the transport change together with OpenAI’s January 6 text-to-speech shutdown. Source Fix
Dec 16, 2026 Passed Azure AI Foundry gpt-image-1.5 retires on Azure. See the Azure retirement schedule Azure image deployments of gpt-image-1.5 stop. Source Fix
Jan 6, 2027 Passed OpenAI Shut down: tts-1, tts-1-hd and gpt-4o-mini-tts (2025-03-20 and 2025-12-15 snapshots). gpt-realtime-2.1-mini The replacement is a Realtime model: a WebSocket session instead of a REST call to /v1/audio/speech. A few lines of code become connection, streaming and reconnect handling. Source Fix
Jan 6, 2027 Passed OpenAI Existing customers can no longer create new fine-tuning jobs. Inference continues until each base model is deprecated. Prompting, retrieval or another provider for new fine-tunes Pipelines that retrain on a schedule stop working. Existing fine-tuned models keep serving until their base model retires. Source Fix
Jan 8, 2027 Passed AWS Bedrock Claude Opus 4.1 reaches end of life (us-east-1, us-east-2, us-west-2). Claude Sonnet 4.5 moves to extended access at higher prices. Newer Claude models Opus 4.1 calls fail. Sonnet 4.5 keeps working but costs more until its end of life on April 8, 2027. Source Fix
Jan 14, 2027 Passed OpenAI Shut down: gpt-4o (2024-08-06 and 2024-11-20), gpt-4o-mini (2024-07-18), gpt-4.1 and gpt-4.1-mini, plus their fine-tunes. gpt-6-luna gpt-6-luna is a reasoning model. Tool calls over Chat Completions need reasoning_effort: none, and temperature and max_tokens can be rejected. Azure keeps non-fine-tuned gpt-4o and gpt-4.1 until April 14, 2027. Source Fix
Jan 20, 2027 Passed OpenAI Legacy realtime and audio models shut down, including gpt-realtime, gpt-4o-realtime, gpt-audio, gpt-4o-audio, gpt-audio-mini, gpt-4o-mini-audio, gpt-realtime-mini, gpt-4o-mini-realtime and gpt-4o-mini-transcribe-2025-03-20. gpt-realtime-2.1, gpt-audio-1.5 or gpt-realtime-2.1-mini; gpt-4o-mini-transcribe-2025-12-15 Voice agents and call bots on these models disconnect. Session settings and voices need re-testing on the replacement. Source Fix
Jan 28, 2027 Passed Google Vertex AI gemini-3.7-flash retires on Vertex AI. gemini-3.8-flash Pinned calls stop resolving. Re-test prompts on the replacement. Source Fix
Jan 31, 2027 Passed LangSmith Old SDK methods and endpoints are removed. The v1 run-sharing endpoints already return a Sunset: 2027-01-31 header. Current SDK and endpoints (LangChain migration guide) Tracing and run-sharing calls from old SDK versions fail. Upgrade the SDK and check any direct API calls. Source Fix
Feb 26, 2027 Passed OpenAI Shut down: whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe, gpt-4o-transcribe-diarize and /v1/audio/translations. See OpenAI's deprecations page Transcription and translation calls fail. Source Fix
Apr 1, 2027 Passed OpenAI Shut down: gpt-5.3-codex, gpt-5.4-nano and gpt-5.1. See OpenAI's deprecations page Calls fail. Source Fix
Apr 8, 2027 Passed AWS Bedrock Claude Sonnet 4.5 reaches end of life on Bedrock. Newer Claude models Calls fail. Source Fix
Apr 14, 2027 Passed Azure AI Foundry gpt-4o and gpt-4.1 (not fine-tuned) retire on Azure. See the Azure retirement schedule Azure deployments stop, three months after OpenAI’s own shutdown of the same models. Source Fix

Three things that make these dates worse than they look

  1. The replacement is often a different kind of model

    GPT-4o and GPT-4.1 are replaced by gpt-6-luna, a reasoning model with different parameters and tool-call behavior. tts-1 is replaced by a Realtime model over WebSocket. Changing the model name is not the migration.

  2. The quick fix often retires next

    Teams fixing the October 23 404 errors moved to gpt-4o or gpt-image-1-mini, which shut down on January 14 and December 1.

  3. The same model retires on different dates on different clouds

    Azure retires o3 on November 19, three weeks before OpenAI does, but keeps GPT-4o until April 14, 2027, three months after OpenAI. Bedrock's dates are different again.

Notes on specific shutdowns

Want it done for you? The migration sprint inventories every call, moves model names into one config, migrates them and replays 50 real prompts before release. $2,500, or $1,500 for a single endpoint.

Find out what breaks before your users do.

A 20-minute call, then a written scope with acceptance criteria and a fixed price within one business day. If it isn't a fit, we'll say so on the call.

Book a 20-minute scoping call See packages and prices

Or email hello@destello.us or call (628) 202-5322. We reply within one business day.