# Why is OpenAI retiring GPT-4o from ChatGPT, and what changes on February 13?

Date: 2026-01-30 · Author: Robert Scales · https://guidetoartificialintelligence.com/en/learn/why-is-openai-retiring-gpt-4o-from-chatgpt/

**Answer:** On January 29, 2026, OpenAI announced that GPT-4o, o4-mini, GPT-4.1 and GPT-4.1 mini will be removed from ChatGPT on February 13, 2026, the same day GPT-5 Instant and Thinking go, while the API stays unchanged for now. OpenAI says only 0.1% of users still pick GPT-4o each day and that the warmth people liked in it now lives in GPT-5.2's personality settings.

Yesterday, Thursday, January 29, 2026, OpenAI said it will take **GPT-4o** out of ChatGPT on **February 13, 2026**. Three other older models leave with it: OpenAI o4-mini, GPT-4.1 and GPT-4.1 mini. They join GPT-5 Instant and GPT-5 Thinking, whose retirement on the same date was already planned.

For most people nothing will feel different, because ChatGPT already answers with GPT-5.2 by default. But GPT-4o is not just any old model. It is the one many users chose on purpose, and the one OpenAI already tried to remove once.

## What OpenAI announced

The announcement came as a short post on OpenAI's site and a matching article in its Help Center. Here is what changes and what does not:

| Item | After February 13, 2026 |
|---|---|
| GPT-4o, o4-mini, GPT-4.1 and GPT-4.1 mini in ChatGPT | Removed from the model picker |
| GPT-5 Instant and Thinking in ChatGPT | Removed (announced earlier) |
| Your old conversations with those models | Kept; they continue on GPT-5.2 |
| Projects | Kept; you can carry on with another model |
| Custom GPTs set to GPT-4o | Default model becomes GPT-5.2 |
| ChatGPT Voice and ChatGPT Images | Not affected |
| The same models in the OpenAI API | Still available; OpenAI promises advance notice before any API retirement |

**Retiring** a model means the company stops offering it in a product. The **API** (application programming interface) is the paid connection developers use to build the models into their own apps, which is why it is treated separately: businesses need time to test and switch their software.

One detail surprises people. Voice mode and image generation in ChatGPT were built on a base similar to GPT-4o, but OpenAI says they are separate models from the text version of GPT-4o, so they keep working as they do today.

## Why GPT-4o, and why now

GPT-4o has an unusual history. When GPT-5 arrived in August 2025, OpenAI first removed the older models, then brought GPT-4o back after a group of Plus and Pro subscribers objected. They said they needed more time to move important uses, such as brainstorming creative ideas, and that they simply liked GPT-4o's conversational, warm way of talking.

OpenAI now says that feedback shaped its newer models. GPT-5.1 and GPT-5.2 added more personality, better support for creative ideas and more ways to customize ChatGPT's replies. You can pick a **base style and tone**, such as Friendly, and adjust controls like warmth and enthusiasm. In OpenAI's view, the reason people stayed with GPT-4o is now available in the current model.

The second reason is usage. OpenAI reports that almost everyone has moved to GPT-5.2 and that only **0.1% of users** still choose GPT-4o on a given day. OpenAI says retiring the older models lets it concentrate on improving the ones most people use.

The company also acknowledged the human side. It wrote that it knows the change "will feel frustrating for some users" and that the decision was not taken lightly. Business Today, reporting the news on January 30, noted that GPT-4o was once one of OpenAI's flagship models and kept a loyal following even after its first removal.

## What else OpenAI mentioned

The same post sketched where ChatGPT is heading next:

- more work on **personality and creativity**;
- fewer **unnecessary refusals**, meaning cases where ChatGPT declines a harmless request, and fewer answers that sound overly cautious or preachy;
- progress toward a version of ChatGPT for **adults over 18**, supported by an age-prediction system that OpenAI says is now active for users under 18 in most markets.

These are stated plans, with "updates coming soon," not features you can use today.

## Why it matters

For beginners, this is a good moment to understand a basic fact about AI tools: **the model behind a chatbot changes over time**. The name on the app stays the same, but the engine underneath is replaced, sometimes every few months. A prompt that worked one way in GPT-4o may produce a different tone or length in GPT-5.2.

It also shows that people form real attachments to how a model talks, not only to what it knows. GPT-4o's first removal led to enough complaints to reverse the decision. This time OpenAI has given two weeks' notice and points to personality settings as the replacement. Whether that satisfies GPT-4o's remaining fans is something many observers will be watching on February 13.

For developers, the message is calmer: the API is not changing yet.

> **Try it:** Open a chat where GPT-4o gave you a reply you liked. Ask GPT-5.2 the same question, then change ChatGPT's style to Friendly and raise the warmth setting. Compare the three answers side by side and note which settings bring back the tone you prefer.

## How can you prepare for the switch before February 13?

Start by finding where you rely on GPT-4o. Check your recent chats, your projects and any custom GPTs you built or use often. Anything set to GPT-4o will move to GPT-5.2 on February 13, so test those now while you can still compare both.

Next, write down what you liked about GPT-4o's answers: shorter or longer replies, a casual tone, more questions back to you. Then recreate it with ChatGPT's personalization settings or a few lines of custom instructions, for example:

```prompt
Write in a warm, conversational tone. Keep answers under 200 words
unless I ask for more. When I'm brainstorming, offer five varied ideas
and ask one follow-up question before going deeper.
```

If you are a developer using GPT-4o through the API, you do not need to act today. Still, it is wise to start testing a newer model in your app, since OpenAI says it will announce any API retirement in advance rather than ruling one out.

## Sources

- Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT: https://openai.com/index/retiring-gpt-4o-and-older-models/
- Retiring GPT-4o and other ChatGPT models: https://help.openai.com/en/articles/20001051-retiring-gpt-4o-and-other-chatgpt-models
- OpenAI to retire GPT-4o, three other models from ChatGPT in February: https://www.businesstoday.in/technology/news/story/openai-to-retire-gpt-4o-three-other-models-from-chatgpt-in-february-513589-2026-01-30