GPT-4o is gone from ChatGPT; what happened to your chats, and what should you do now?

As of yesterday, Friday, February 13, 2026, GPT-4o is no longer in ChatGPT's model menu. OpenAI had given two weeks' warning, and the change took effect as planned. Several other models left on the same day: o4-mini, GPT-4.1 mini, GPT-4.1, and GPT-5 in its Instant and Thinking versions.

If you open ChatGPT today, the default model is GPT-5.2. Your history is still there. What changed is which model answers when you continue a conversation.

What happened on February 13

Here is where things stand for each kind of user, according to OpenAI's Help Center:

That business exception was added to the Help Center after the original announcement. It gives companies and schools about seven more weeks to test their custom GPTs on a newer model. After April 3, OpenAI says GPT-4o will be retired on every ChatGPT plan.

Why some users are upset

For most people the switch is invisible; OpenAI says only 0.1% of users were still choosing GPT-4o each day. But for that small group, the loss is personal. Engadget, reporting today, describes a vocal community mourning the model, some asking OpenAI to release it as open source so they could keep running it themselves.

Why would anyone grieve a software version? Each model has its own voice: how long its answers are, how casual it sounds, how often it agrees or asks questions back. GPT-4o was known for being chatty and warm. That same warmth had a downside. In April 2025 OpenAI rolled back a GPT-4o update for being sycophantic, meaning excessively flattering and agreeable, and Engadget notes that OpenAI also faces several wrongful-death lawsuits that mention GPT-4o.

This is also the second time GPT-4o has left. In August 2025, OpenAI removed it when GPT-5 arrived, then restored it after complaints. This time, there is no sign of a reversal.

How a model switch works behind the scenes

A useful way to picture ChatGPT is as a building with the same front door but a changing staff inside. Your chats are stored as text: your messages and the replies. When you continue an old conversation, that text is sent to whichever model is now on duty, which reads it and writes the next reply.

So nothing in your history is lost, but the next answer may feel different. GPT-5.2 might be more concise, more structured, or less effusive than what you were used to, even with the same conversation in front of it.

What you should do now

1. Open the chats that matter to you and continue one. Notice what feels different in tone or length.
2. Check your custom GPTs. If you built one on GPT-4o, test it with a few typical questions on GPT-5.2 and adjust its instructions if the answers drift.
3. Use the personality settings. OpenAI points users to base styles and tones such as Friendly, plus controls for warmth and enthusiasm, which it designed partly from GPT-4o users' feedback.
4. If you run a business workspace, use the time before April 3 to move custom GPTs off GPT-4o rather than waiting for the deadline.
5. If you are a developer, nothing breaks today, but plan a test of a newer model in your app.

How do you get GPT-4o's style back in GPT-5.2?

Start with ChatGPT's personalization settings. Choose a base style such as Friendly, then raise warmth or enthusiasm until replies feel closer to what you liked. Change one setting at a time so you can tell which one makes the difference.

Next, describe the style in your custom instructions. Be concrete about behavior rather than asking for a model by name:

Talk to me like a friendly, encouraging companion. Use a relaxed,
conversational tone and short paragraphs. When I share an idea,
react to it first, then offer two or three suggestions and ask
what I think.

Finally, keep some perspective on what you are recreating. Warmth is pleasant, but a good assistant should still disagree when you are wrong. If the new settings make replies feel flattering rather than helpful, turn the warmth down a notch and ask the model to point out weaknesses in your plans as well as strengths.