Which AI models are being shut down, and when?

Announced API shutdown dates for GPT-4o, GPT-4.1, Claude, Gemini and other models, with the suggested replacement and a source for each.

Free. Runs in your browser. No sign-up.

A shutdown forces a switch. To switch without your users noticing, run the replacement on real requests from your app before the date, and check its results work as often as the old model's. The model switch test does that side by side.

Show shutdowns from

Prices and dates checked 2026-11-12. Days are counted from that date, 2026-11-12.

Upcoming (21)

2026-11-30
18 days left
Anthropic
claude-sonnet-4-5-20250929
Replacement: claude-sonnet-5-5
Test the replacement on your real requests Source: platform.claude.com
2026-12-01
19 days left
OpenAI
gpt-image-1.5, gpt-image-1-mini, chatgpt-image-latest
Replacement: gpt-image-2.6-sunburst or gpt-image-2.6-flare
Test the replacement on your real requests Source: developers.openai.com
2026-12-01
19 days left
Google
gemini-2.5-pro
Replacement: gemini-3.8-flash (requests auto-route, billed at 3.8 Flash rates)
Test the replacement on your real requests Source: aistudio.google.com
2026-12-01
19 days left
Google
gemini-2.5-flash
Replacement: gemini-3.5-flash-lite (requests auto-route, billed at 3.5 Flash-Lite rates)
Test the replacement on your real requests Source: aistudio.google.com
2026-12-03
21 days left
Google
gemini-omni-1.1-flash
Replacement: gemini-omni-1.2-flash (requests auto-route at 1.2 rates)
Test the replacement on your real requests Source: aistudio.google.com
2026-12-11
29 days left
OpenAI
gpt-5, o3
Replacement: gpt-5.6-sol
Test the replacement on your real requests Source: developers.openai.com
2026-12-11
29 days left
OpenAI
gpt-5-mini
Replacement: gpt-5.6-terra
Test the replacement on your real requests Source: developers.openai.com
2026-12-11
29 days left
OpenAI
gpt-5-nano
Replacement: gpt-5.6-luna
Test the replacement on your real requests Source: developers.openai.com
2026-12-11
29 days left
OpenAI
gpt-5-pro, o3-pro
Replacement: gpt-5.6-sol with reasoning.mode: pro
Test the replacement on your real requests Source: developers.openai.com
2026-12-31
49 days left
Moonshot AI
Kimi K2 series (kimi-k2.6, kimi-k2.7-code)
Replacement: kimi-k3.1-pro
Test the replacement on your real requests Source: platform.kimi.ai
2027-01-06
55 days left
OpenAI
tts-1, tts-1-hd, gpt-4o-mini-tts
Replacement: gpt-realtime-2.1-mini
Test the replacement on your real requests Source: developers.openai.com
2027-01-14
63 days left
OpenAI
gpt-4o (gpt-4o-2024-08-06, gpt-4o-2024-11-20)
Replacement: gpt-6-luna
Test the replacement on your real requests Source: developers.openai.com
2027-01-14
63 days left
OpenAI
gpt-4o-mini (gpt-4o-mini-2024-07-18)
Replacement: gpt-6-luna
Test the replacement on your real requests Source: developers.openai.com
2027-01-14
63 days left
OpenAI
gpt-4.1 (gpt-4.1-2025-04-14)
Replacement: gpt-6-luna
Test the replacement on your real requests Source: developers.openai.com
2027-01-14
63 days left
OpenAI
gpt-4.1-mini (gpt-4.1-mini-2025-04-14)
Replacement: gpt-6-luna
Test the replacement on your real requests Source: developers.openai.com
2027-02-26
106 days left
OpenAI
whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe
Replacement: gpt-transcribe or gpt-live-transcribe
Test the replacement on your real requests Source: developers.openai.com
2027-03-15
123 days left
Google
gemini-2.5-flash-image (Nano Banana)
Replacement: gemini-3.1-flash-lite-image
Test the replacement on your real requests Source: aistudio.google.com
2027-04-01
140 days left
OpenAI
gpt-5.1, gpt-5.3-codex
Replacement: gpt-6-sol
Test the replacement on your real requests Source: developers.openai.com
2027-04-01
140 days left
OpenAI
gpt-5.4-nano
Replacement: gpt-6-luna
Test the replacement on your real requests Source: developers.openai.com
2027-04-30
169 days left
Mistral
mistral-medium-3-5 (Mistral Medium 3.5)
Replacement: Mistral Large 4
Test the replacement on your real requests Source: docs.mistral.ai
2027-04-30
169 days left
Mistral
mistral-large-2512 (Mistral Large 3)
Replacement: Mistral Large 4
Test the replacement on your real requests Source: docs.mistral.ai

Already shut down (11)

2026-10-23
20 days ago
OpenAI
gpt-image-1
Replacement: gpt-image-2.5-sunburst or gpt-image-2.5-flare
Source: developers.openai.com
2026-10-23
20 days ago
OpenAI
gpt-4.1-nano, o4-mini, o3-mini, o1, gpt-4-turbo
Replacement: gpt-5.6-sol or gpt-5.6-terra
Source: developers.openai.com
2026-10-22
21 days ago
Google
veo-3.1-generate-preview, veo-3.1-fast-generate-preview, veo-3.1-lite-generate-preview (Gemini API)
Replacement: gemini-omni-1.2-flash (or GA Veo on Google Cloud)
Source: aistudio.google.com
2026-09-24
49 days ago
OpenAI
Sora 2 / sora-2-pro and the Videos API
Replacement: None named by OpenAI
Source: developers.openai.com
2026-08-26
78 days ago
OpenAI
Assistants API
Replacement: Responses API and Conversations API
Source: developers.openai.com
2026-08-17
87 days ago
Google
imagen-4.0-generate-001, imagen-4.0-ultra-generate-001, imagen-4.0-fast-generate-001 (Gemini API)
Replacement: gemini-nano-banana-2.1
Source: aistudio.google.com
2026-08-05
99 days ago
Anthropic
claude-opus-4-1-20250805
Replacement: claude-opus-4-8
Source: platform.claude.com
2026-06-15
150 days ago
Anthropic
claude-opus-4-20250514
Replacement: claude-opus-4-8
Source: platform.claude.com
2026-06-15
150 days ago
Anthropic
claude-sonnet-4-20250514
Replacement: claude-sonnet-5-5
Source: platform.claude.com
2026-06-01
164 days ago
Google
gemini-2.0-flash
Replacement: gemini-3.6-flash
Source: aistudio.google.com
2026-02-19
266 days ago
Anthropic
claude-3-5-haiku-20241022
Replacement: claude-haiku-4-5-20251001
Source: platform.claude.com

What a shutdown means for your API calls

On the shutdown date the provider stops serving the model. Requests that name it start returning errors, so any feature built on it stops working until you change the model id. Some providers route old ids to a newer model instead, which keeps calls working but changes the results and the price without any change on your side. Either way, the date is a deadline for a switch you should make on purpose, not one that gets made for you.

How to switch without your users noticing

Find every place your app names the model, including background jobs, scripts and any judge or moderation step.
Run real past requests on the replacement, including ones where the old model's result didn't work, and read the two results side by side. The model switch test does this for your own prompt, and each “Test the replacement” link above picks both models for you when we know them.
Check the real cost, not just the price per token. A replacement that writes more, or whose results fail more often, can cost more for each result that worked.
Switch behind a flag for a share of users, compare how often results work first time on the old model and the new one, then move everyone over well before the date. The A/B sample size calculator says how many results that comparison needs.
Leave the old model id on the results it made, so your history still says what produced each one.

The GPT-4o and GPT-4.1 shutdown on 2027-01-14

OpenAI has set 14 January 2027 as the API shutdown date for gpt-4o, gpt-4o-mini, gpt-4.1 and gpt-4.1-mini, and names gpt-6-luna as the replacement for all four. GPT-6 Luna lists at $0.10 per million input tokens and $0.50 per million output tokens, against $2.50 and $10.00 for GPT-4o, so the price per call should fall.

The price is the easy part. Prompts tuned on GPT-4o can behave differently on a new model family: length, formatting, tone and refusals can all shift, and a support bot or an invoice reader that worked for most requests can start failing for some of them. If your app was built on GPT-4o or GPT-4.1, this is the shutdown to plan for now. Test GPT-6 Luna against GPT-4o on your own prompt early, so you have weeks to fix your prompt rather than days.

Questions

What happens when a model is shut down?

API calls that name it start failing. Switch to the replacement before the date, and test it on your own requests first, because its results will differ.

How do I switch without my users noticing?

Run the replacement on real requests from your app, including some that went badly before, and check its results work at least as often as the old model's. The model switch test does this side by side with your own provider key, and each entry here links to it with both models picked.

How current is this list?

Each entry links to its source and the page shows the date it was last checked.