OpenAI will remove four GPT-4 series models from its API on 14 January 2027: GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini. It told developers using them on 14 October 2026 and names GPT-6 Luna as the recommended replacement. If a tool your business relies on was built on one of these models, it has to be moved before that date or it will stop working.
What happened
On 14 October 2026 OpenAI added a new entry to its API deprecations page, headed "GPT-4 series model deprecations". The entry says OpenAI notified developers using certain GPT-4 series models of their deprecation and removal from the API on 14 January 2027. OpenAI says that when it announces a deprecation it emails the customers who are actively using the model and records the change on that page.
In OpenAI's terms, a model is deprecated as soon as the deprecation is announced, and every deprecated model has a shut down date. After that date the model is no longer accessible. The page uses "sunset" and "shut down" to mean the same thing.
The API (application programming interface) is the service developers use to connect their own software to OpenAI's models. The ChatGPT app is a separate product. OpenAI's Help Center says GPT-4o, GPT-4.1 and GPT-4.1 mini were retired in ChatGPT on 13 February 2026, and that these models would continue to be available in the API. The notice of 14 October 2026 now sets an end date for the API as well.
Key details
| Model | API model names | Shutdown date | Recommended replacement |
|---|---|---|---|
| GPT-4o | gpt-4o-2024-08-06, gpt-4o-2024-11-20 | 14 January 2027 | gpt-6-luna |
| GPT-4o mini | gpt-4o-mini-2024-07-18 | 14 January 2027 | gpt-6-luna |
| GPT-4.1 | gpt-4.1-2025-04-14 | 14 January 2027 | gpt-6-luna |
| GPT-4.1 mini | gpt-4.1-mini-2025-04-14 | 14 January 2027 | gpt-6-luna |
- Fine-tuned versions of these models (versions trained further on a customer's own examples) follow the same shutdown schedule as the base models they were built on.
- The same page says that from 6 January 2027 active existing customers can no longer create new fine-tuning jobs.
- OpenAI says that in some cases developers may be able to provision dedicated capacity for continued access after a model's shutdown date, by contacting its sales team.
- OpenAI's GPT-6 Luna model page describes the replacement as its most efficient model for focused, high-volume tasks. It takes text and image input, gives text output, and has a standard price of US$0.10 per million input tokens and US$0.50 per million output tokens. Tokens are the small pieces of text a model reads and writes, and API use is billed by them.
Why it matters
Many business tools are not built by the business that uses them. A chat assistant on your website, a plug-in that drafts quotes, an automation that sorts incoming enquiries or a reporting tool may call one of these models in the background, set up by a developer or a software vendor some time ago. The people using the tool every day may not know which model sits behind it.
Once the shutdown date passes, the retired model is no longer accessible to the software that calls it. OpenAI's own page says software relying on its models may need occasional updates to keep working. For a business, the risk is finding out on the day: a chat assistant that goes quiet, or an automation that fails without telling anyone.
The timing is worth noting. OpenAI's deprecations page sets minimum notice periods, unless safety or compliance concerns require a faster timeline: at least 6 months for generally available models, at least 3 months for specialised variants, and possibly as little as 2 weeks for preview models, which it does not recommend for business-critical work. The entry of 14 October 2026 leaves three months between notice and shutdown and does not give a reason for its timeline. Either way, the date is set and the planning has to start now.
Moving to a new model is also not a simple name swap. GPT-6 Luna supports reasoning effort settings, with medium as the default. Its model page says Chat Completions, an older way of calling the API, supports function calling with GPT-6 Luna only when reasoning_effort is set to none, and points developers to the Responses API for built-in tools and function calling. Function calling is how a model asks your software to run an action, such as looking up an order. A tool written against the older setup may need more than a one-line change, and its answers, speed and running cost can all shift, so the move needs testing.
What this means for businesses
Most small and medium businesses do not call OpenAI's API themselves. The question is whether something you pay for, or had built, does. A short checklist:
- List every tool in the business that uses AI: website chat, email or ticket sorting, quote and document drafting, internal assistants, and any custom software or integration.
- Ask each vendor or developer in writing whether the tool uses GPT-4o, GPT-4o mini, GPT-4.1 or GPT-4.1 mini through OpenAI's API, and when they will move it before 14 January 2027.
- If the business has its own OpenAI API account, check that the email address on it reaches someone who reads it, because OpenAI sends deprecation notices by email to customers actively using a model.
- Ask your developer to compare the model names in your code and settings with the table above. OpenAI also says the List models endpoint (GET /v1/models) has an optional shutdown_date field that software can check automatically.
- If you use a fine-tuned model based on one of these, plan its replacement first: it shuts down with its base model, and from 6 January 2027 active existing customers can no longer create new fine-tuning jobs.
- Test the replacement on real examples of your work before switching: the quality of the answers, the format your other systems expect, the speed and the monthly cost.
- Name one person who owns the change and write down the date it will be finished.
If the AI feature sits inside a larger system, such as a CRM or a quoting tool connected to your accounting software, treat the model change as part of the review of your business systems and integrations. For software built for you, check who holds the code and the API account: when we build custom software, clients own their code, data and documentation. If you are not sure which tools are affected, IT consulting can start with an inventory of what you use. For the latest changes inside ChatGPT itself, see our brief on GPT-6 and Intelligent UI in ChatGPT.
We can help you find out which of your tools depend on these models and plan the move. The first consultation is free, and you receive a written quote before any work starts: ask us for a quote.
Frequently asked questions
When will OpenAI remove GPT-4o from its API?
OpenAI's deprecations page gives 14 January 2027 as the shutdown date for GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini in the API. After that date the models are no longer accessible.
Does this notice change the ChatGPT app?
This notice is about the API, the service other software uses to connect to OpenAI's models. OpenAI's Help Center says GPT-4o, GPT-4.1 and GPT-4.1 mini were already retired in ChatGPT on 13 February 2026.
What does OpenAI recommend instead?
OpenAI lists gpt-6-luna as the recommended replacement for every model in this notice. Its model page describes GPT-6 Luna as OpenAI's most efficient model for focused, high-volume tasks.
What happens to fine-tuned models built on GPT-4o or GPT-4.1?
OpenAI says fine-tuned versions follow the same shutdown schedule as their base models. Its deprecations page also says active existing customers can no longer create new fine-tuning jobs from 6 January 2027.
How do I know whether my business is affected?
Ask every vendor and developer whose tools use AI whether they call these models through OpenAI's API, and check the email address on any OpenAI API account you hold. A developer can also compare the model names in your code with OpenAI's list.
Key takeaways
- OpenAI will remove GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini from its API on 14 January 2027.
- Developers were notified on 14 October 2026, and OpenAI recommends gpt-6-luna as the replacement.
- Fine-tuned versions shut down with their base models, and from 6 January 2027 existing customers can no longer start new fine-tuning jobs.
- Ask your vendors and developers now which of your tools use these models.
- Test the replacement on real work before switching, and give one person ownership of the change.