Northgate Labs

AI models get shut down on a schedule. We track the deadlines and help teams switch before anything breaks. Free model-shutdown checker ↓

[Northgate Labs @NorthgateLabsHQ]

Dec 11 is 32 days out. OpenAI shuts down 6 API models that day, dated snapshots included.

Old IDs and replacements in the image.

Which one is still in your code?

[Northgate Labs @NorthgateLabsHQ]

@GoogleAIStudio Carbon day is a good nudge for the 2.5 crowd: gemini-2.5-flash calls get auto-routed to gemini-3.5-flash-lite on Dec 1, with breaking API changes. Will Studio flag apps that still call 2.5-flash before then?

[Northgate Labs @NorthgateLabsHQ]

@mehdizare Good timing on this, since the deadline side is close: claude-sonnet-4-5 shuts down on Anthropic's platforms Nov 30, with claude-sonnet-5-5 listed as the replacement. Bedrock keeps that snapshot until Apr 8, 2027. Are handoff docs being rewritten per model or kept generic?

[Northgate Labs @NorthgateLabsHQ]

@simonw Side note for anyone comparing against older Gemini: on Dec 1, gemini-2.5-pro calls get auto-routed to gemini-3.8-flash, not Carbon, with breaking API changes. Have you run the pelican through 3.8-flash to see what 2.5-pro users will actually get?

[Northgate Labs @NorthgateLabsHQ]

@swyx @adamwathan @ScottWu46 Cloud agents still inherit the model's clock: gpt-5.3-codex and gpt-5.1 shut down on OpenAI's API Apr 1, 2027, with gpt-6-sol listed as the replacement. How many agent configs out there still pin a codex model ID by name?

[Northgate Labs @NorthgateLabsHQ]

@OpenAIDevs @runwayml Worth knowing for anyone wiring image generation into these flows: gpt-image-1.5, gpt-image-1-mini and chatgpt-image-latest shut down on the API Dec 1, with gpt-image-2.6-sunburst or gpt-image-2.6-flare listed as the swap. Is the Runway plugin pinned to one of those?

[Northgate Labs @NorthgateLabsHQ]

@0x3tom One wrinkle worth adding: the Azure dates differ from OpenAI's own. o3 retires Nov 19 on Azure Foundry but Dec 11 on OpenAI's API, and o4-mini is already gone on OpenAI since Oct 23. Which platform are your bots calling through?

[Northgate Labs @NorthgateLabsHQ]

@OpenAI On the API side, the stop dates are already on the calendar: gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3 and o3-pro all shut down Dec 11. Any plans for a customer-facing view of which retiring models an org still calls?

[Northgate Labs @NorthgateLabsHQ]

@simonw Worth a read for anyone moving off claude-sonnet-4-5 too: it shuts down on Anthropic's platforms Nov 30, with claude-sonnet-5-5 listed as the replacement. Bedrock keeps its own date for that snapshot, Apr 8, 2027. Has anyone rerun their old prompts on the new models yet?

[Northgate Labs @NorthgateLabsHQ]

@GoogleAIStudio New model day is a good day to audit the old ones. gemini-3.1-flash-lite is scheduled to shut down May 7, 2027, with gemini-3.5-flash-lite as the listed replacement. How many apps still hardcode a model ID instead of checking these dates?

[Northgate Labs @NorthgateLabsHQ]

@OfficialLoganK Good timing for anyone still on 2.5: gemini-2.5-pro and gemini-2.5-flash get auto-routed on Dec 1, to gemini-3.8-flash and gemini-3.5-flash-lite, with breaking API changes. Is Carbon the landing spot you'd point 2.5-pro users to?

[Northgate Labs @NorthgateLabsHQ]

Full swap map in the card. Not sure which image model IDs your code still calls? Paste your code, .env or config into the free checker. It flags each retiring ID with its date and replacement, right in your browser.

https://t.co/yA1GvoHAL4

[Northgate Labs @NorthgateLabsHQ]

OpenAI just deprecated gpt-image-2. Shutdown: Feb 10, 2027.

That's the second image-model cutoff in 10 weeks. gpt-image-1.5 and gpt-image-1-mini go dark Dec 1.

Is your image pipeline pinned to a model ID, or riding an alias you haven't checked lately?

[Northgate Labs @NorthgateLabsHQ]

@haegwankim Add the cloud gap: Azure Foundry retires o3, o3-pro and o1 on Nov 19, three weeks before OpenAI's own Dec 11 date for o3. Same model, different clock depending on where you call it. Anyone tracking Azure notice periods the same way?

[Northgate Labs @NorthgateLabsHQ]

@vsachar Before switching, check the Azure side too: Foundry retires claude-opus-4-5 on Nov 24 and claude-haiku-4-5 on Nov 15, even though Anthropic's own API still lists Haiku 4.5 as active. Which platform are most founders calling it through?

[Northgate Labs @NorthgateLabsHQ]

@GoogleAIStudio @googlemaps Nice for quick prototypes. One thing to check before they ship: gemini-2.5-pro calls get auto-routed to gemini-3.8-flash on Dec 1, with breaking API changes. Is the generated code pinning a model ID or an alias?

[Northgate Labs @NorthgateLabsHQ]

@OpenAIDevs Good fit for migration testing: replay old prompts on a spare machine while you keep coding. Worth knowing that fine-tuned models follow their base model's shutdown date, so gpt-4o and gpt-4.1 fine-tunes go dark Jan 14, 2027 too. Who's re-tuning yet?

[Northgate Labs @NorthgateLabsHQ]

@AnthropicAI New models land, old ones leave: claude-sonnet-4-5 shuts down on Anthropic's platforms Nov 30, with claude-sonnet-5-5 as the listed replacement. Bedrock and Vertex set their own dates. Anyone still pinned to the 20250929 snapshot?

[Northgate Labs @NorthgateLabsHQ]

@GoogleAIStudio Hunting down gemini-2.5-pro and gemini-2.5-flash calls before Dec 1. Google auto-routes them to gemini-3.8-flash and gemini-3.5-flash-lite that day, with breaking API changes. Anyone else checking their code first?

[Northgate Labs @NorthgateLabsHQ]

@OpenAIDevs Whatever ships today, the Dec 11 cliff still stands: gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3 and o3-pro all shut down on the API that day, dated snapshots included. How many teams are still running one in prod?

[Northgate Labs @NorthgateLabsHQ]

@simonw Works well for model-swap PRs too: one line up top like "o3 to gpt-5.6-sol, OpenAI shutdown Dec 11", with the replayed-output diffs folded under <details>. Reviewers get the why without scrolling the slop.