AI model shutdowns: check your workflows before 14 January 2027

Short answer: OpenAI shuts down the gpt-4o, gpt-4o-mini, gpt-4.1 and gpt-4.1-mini snapshots, and every fine-tune built on them, on 14 January 2027. Its replacement is gpt-6-luna. Before that, claude-sonnet-4-5-20250929 retires on 30 November 2026 and gpt-5, gpt-5-mini and o3 on 11 December 2026. Paste a workflow, code or config below to see which models you use and when each one stops working.

Check your workflows and code

Shutdown calendar

OpenAI's 14 January notice names the dated snapshots (for example gpt-4o-mini-2024-07-18). The aliases gpt-4o-mini and the others point to those snapshots, and OpenAI's model pages mark them deprecated, but the notice does not list them by name. Plan as if they stop on the same day.

Fix it where the model is set

• n8n: n8n saves the model ID inside every AI node, and several of n8n's own defaults are on the shutdown list: the OpenAI Chat Model defaults to gpt-5-mini (shuts down 11 December 2026), Anthropic Chat Model 1.3 defaults to claude-sonnet-4-5-20250929 (30 November 2026), and the OpenAI node's transcribe and translate operations hardcode whisper-1 (26 February 2027).
• WordPress AI plugins: WordPress AI plugins keep the OpenAI model in their own settings, and some fall back to a retiring model when the setting is empty: LocoAI uses gpt-4o-mini, and Uncanny Automator's OpenAI actions request gpt-4, which OpenAI shut down on 23 October 2026.
• Make.com: Make moved scenarios onto gpt-4o-2024-08-06 in February 2026 and onto gpt-4.1 in March 2026 when older OpenAI models were retired.
• Zapier: Zapier has no general help article on retired OpenAI models, and its Send Prompt troubleshooting article still suggests babbage-002, davinci-002 and gpt-3.5-turbo-instruct, which OpenAI shut down on 28 September 2026.
• Python, Node.js and LangChain code: Search your code, environment files and config for the retiring model IDs and change them before each date: gpt-4o, gpt-4o-mini, gpt-4.1 and gpt-4.1-mini snapshots stop on 14 January 2027 (replacement gpt-6-luna), claude-sonnet-4-5-20250929 on 30 November 2026.

gpt-6-luna is not a drop-in swap

• Tool calls on Chat Completions: OpenAI says Chat Completions supports function calling with gpt-6-luna "only with reasoning_effort set to none". The default effort is medium, so agents that call tools through Chat Completions need reasoning_effort: "none" or the Responses API.
• Sampling settings: when reasoning effort is not none, remove temperature, top_p and top_logprobs (and logprobs on Chat Completions).
• Price: $0.10 per million input tokens and $0.50 per million output tokens, against $0.15 and $0.60 for gpt-4o-mini. Reasoning tokens are billed as output, so compare a real day of traffic before assuming it is cheaper.
• Claude: Anthropic API version 2026-10-21 returns a 400 error if a request sends temperature, top_p or top_k, "even at the parameter's default value". The new major SDKs (Python anthropic 2.0.0, @anthropic-ai/sdk 1.0.0) switch to that version automatically.

What the error looks like

OpenAI does not document the error for a shut-down model. The day after its July 2026 shutdown, developers saw Chat Completions answer:

The model `computer-use-preview-2025-03-11` has been deprecated

while the Responses API answered "Model not found computer-use-preview-2025-03-11". A model ID OpenAI does not recognise returns 404 model_not_found ("The model … does not exist or you do not have access to it"). Any of these after a deadline means the model ID has to change. See the fix page for these errors.

Questions

Does the alias gpt-4o-mini stop working on 14 January 2027?

OpenAI's notice lists the snapshot gpt-4o-mini-2024-07-18, which is the only snapshot the alias points to, and the model page marks gpt-4o-mini as deprecated. The notice does not name the alias itself. Treat 14 January as the last day for both and move to a current model before then.

What happens to my fine-tuned models?

OpenAI: "Fine-tuned versions of these models follow the same shutdown schedule." A fine-tune of gpt-4o-mini-2024-07-18 stops on 14 January 2027. New fine-tuning jobs can no longer be created from 6 January 2027.

Can I check this from code?

Yes. OpenAI's List models endpoint has an optional shutdown_date field. The scanner's data is also at /model-registry.json if you want to check a repository in CI.

Is anything uploaded when I use the scanner?

No. The scan runs in your browser and the page makes no network request with what you paste. Remove API keys before pasting anyway.

Sources

• OpenAI: Deprecations (dates, snapshots and replacements)
• OpenAI: gpt-6-luna model page (reasoning effort, pricing, function calling)
• OpenAI: Using the latest model (sampling parameters)
• Anthropic: Model deprecations
• Anthropic: Migrating to API version 2026-10-21
• Google: Gemini API deprecations

Facts checked 28 October 2026 against the providers' own deprecation pages. Not affiliated with OpenAI, Anthropic or Google.