OpenAI switches off GPT-4o and GPT-4.1 in its API on 14 January. Check what your tools run on

On 14 October OpenAI told developers that four widely used models will leave its API on 14 January 2027: GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini. The notice names five dated versions: gpt-4o-2024-08-06, gpt-4o-2024-11-20, gpt-4o-mini-2024-07-18, gpt-4.1-2025-04-14 and gpt-4.1-mini-2025-04-14. Fine-tuned models built on them shut down on the same day. OpenAI's documentation says a model that has been shut down is no longer accessible, so any tool still calling one of these names will stop answering.

OpenAI keeps these dates. On 23 October, two days ago, it shut down an earlier batch it had announced in April, including the first GPT-4o version from May 2024 and GPT-4.1 nano.

Who this affects

The notice is about the API, the layer that sits behind customer service bots, document processing scripts, CRM add-ons and the internal tools an agency or a developer built for you in 2024 or 2025. Those were often set up once with a fixed model name and never touched again. If nobody in the company knows which model a tool calls, that is the first thing to find out.

The replacement is cheaper, with two catches

OpenAI recommends gpt-6-luna for all four models. Its price list puts Luna at $0.10 per million input tokens and $0.50 per million output tokens. GPT-4o costs $2.50 and $10.00, and GPT-4.1 costs $2.00 and $8.00. A tool moving from GPT-4o pays a fraction of what it pays now for the same volume. OpenAI describes Luna as its most efficient model for focused, high-volume tasks, so test the quality of the answers on your own data before you switch.

The first catch: Luna supports Chat Completions, the Responses API and Batch, but not the Assistants API or fine-tuning. A tool built on Assistants, or one that depends on a fine-tuned GPT-4o, needs more than a new model name. The second catch matters for companies that keep data in the EU. Luna is available with EU data residency, and OpenAI charges a 10% surcharge on regional endpoints for models released from 5 March 2026 onwards. Luna falls into that group.

This notice covers OpenAI's own API. If you reach GPT-4o through Microsoft Azure, check Microsoft's retirement dates separately.

What to do on Monday

• ✳List every tool, script and integration that calls the OpenAI API, including ones a partner built for you. Ask suppliers in writing which model their product uses
• ✳Search code and configuration for the strings gpt-4o and gpt-4.1. Each hit is a place that needs a decision before 14 January
• ✳Pick ten to twenty real tasks per tool, run them on gpt-6-luna and compare the answers with what you get today. Do it in November, not in the week before the deadline
• ✳If a tool uses the Assistants API or a fine-tuned model, budget developer time now. Those cases need rework, not a one-line change
• ✳Update your cost forecast. The per-token price drops a lot, but add the 10% surcharge if you use EU data residency

A model shutdown is a maintenance date like a certificate expiry. Companies that keep a list of their AI integrations handle it in an afternoon. Companies that do not find out from a customer in January.