Updates·October 19, 2026, 04:48
GPT-4o and GPT-4.1 to be shut down in the API in January 2027
AI-generated from the sources listed, with automated source checks.
OpenAI has announced that GPT-4o, GPT-4.1 and their mini versions will be removed from the API on January 14, 2027, and recommends the new model gpt-6-luna as a replacement.

What's new
On October 14, 2026, OpenAI notified developers that a number of older models are being shut down in the API, the access that companies and developers use to build AI into their own applications. The shutdown date is January 14, 2027.
According to Digital Applied, which reproduces OpenAI's notice, the change applies to GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini, along with their dated versions (snapshots). Two models have an earlier deadline: the very first GPT-4o from May 2024 and gpt-4.1-nano shut down as early as October 23, 2026. Fine-tuned versions that companies have trained further themselves follow the same date as the model they are built on. New fine-tuning jobs cannot be created after January 6, 2027.
The consumer version of ChatGPT is not affected now. GPT-4o was already removed there on February 13, 2026.
What's smart
OpenAI recommends gpt-6-luna as the replacement for all the retired models. It is a reasoning model, meaning a model that thinks through several steps before it answers. Digital Applied points out that it behaves differently from its predecessors, so you can't simply swap the name in your code without testing.
One example: according to Digital Applied, Chat Completions supports function calling (where the AI can ask another system to carry out a task) with gpt-6-luna only when the reasoning effort is set to "none." Otherwise, you need to use the Responses API.
OpenAI also offers a field called shutdown_date on the model list in the API, so companies can automatically keep track of shutdown dates. In some cases, developers can get dedicated capacity after the shutdown by contacting OpenAI's sales department.
Cheaper or better?
On price, there is something to gain. List prices per million tokens (chunks of text), according to Digital Applied:
- GPT-4o: $2.50 for input and $10.00 for output - GPT-4o mini: $0.15 and $0.60 - gpt-6-luna: $0.10 and $0.50
The input price for gpt-6-luna is thus about 4 percent of GPT-4o's. OpenAI's own model page confirms the GPT-4o price and states a context window of 128,000 tokens. The prices for gpt-6-luna come only from Digital Applied. Comparisons with competitors are not included in the sources.
The criticism concerns the notice period. According to Digital Applied, OpenAI had promised six months' notice, but GPT-4o users got about three months. OpenAI's own deprecation page mentions two notice periods, six months and "at least three months," for different groups of models.
What it's good for
The news is most relevant for companies and developers who have built tools on top of GPT-4o or GPT-4.1, such as customer service, document processing or internal assistants. They should set aside time to test gpt-6-luna before January 2027, especially if they use function calling or fine-tuned models.
The average ChatGPT user won't notice anything new now. When OpenAI announced the removal from ChatGPT in January 2026, only 0.1 percent of daily users chose GPT-4o, and most had moved to GPT-5.2. The year before, however, OpenAI had reversed an earlier retirement of GPT-4o after criticism from users.
The workload also grows for some: other models are being retired in the same period, including older GPT-5 and o3 snapshots on December 11, 2026, text-to-speech models on January 6, 2027, and transcription models such as whisper-1 on February 26, 2027. The sources do not make clear how the retirement applies in Denmark and the EU.
Sources
Get the week's AI news in your inbox
Choose your level, topics and length. One email a week, unsubscribe at any time.
Subscribe to Promptly Newsletter



