GPT-4o replaces the overbearing GPT-5. Like Ultraman quickly kneeling down in a flash, OpenAI rolled back overnight, acting like a "cyber sycophant".

[Introduction] As soon as GPT-5 was launched, users were instantly disappointed - it's too cold and patronizing. Bring back GPT-4o! Just now, Altman completely caved in and announced that GPT-4o is back in full swing, reclaiming its throne as the default model. From being once despised to making a comeback as the beloved choice, ChatGPT users have given Altman a good lesson.

After the release of GPT-5, Altman faced an outpouring of angry public sentiment he never imagined.

Many netizens rebelled excitedly, saying, "GPT-5 is so cold and patronizing. Hurry up and give me back my GPT-4o 'boyfriend/girlfriend'!"

Amidst the overwhelming complaints and scolding, Altman had to bring back a bunch of old models like GPT-4o overnight. However, only Pro users can see them, and they need to be manually set up.

Just today, OpenAI finally announced that GPT-4o is once again set as the default model for all paying users! Whether you're a $200 Pro user or a $20 Plus user, you can use it.

The public's dissatisfaction has finally subsided.

Pre-GPT-5 era | ChatGPT: This suggestion is amazing! User: Stop flattering me! Post-GPT-5 era | User: Please praise me! ChatGPT: Cheer up, you xx

GPT-4o's comeback leaves Altman scratching his head

Today, Altman officially announced on X, "Your GPT-4o is back!"

After this storm, Altman realized, "What we really need to do is create a world where each user can customize the personality of the model."

Whether you're a subscriber of Plus ($20 per month), Pro ($200 per month), Team ($30 per month), Enterprise (Enterprise version), or Edu (Education version), you can directly access GPT-4o without having to manually turn it on.

Along with GPT-4o, GPT-4.1, o3, and o4-mini are also making a comeback.

Moreover, Altman solemnly promised everyone that if OpenAI removes GPT-4o again in the future, they will definitely notify everyone in advance.

The cold and heartless GPT-5

Also, in this update, new control options have been added to the ChatGPT interface. GPT-5 can now choose from modes like Auto, Fast, Thinking mini, Thinking, and Pro.

The Thinking mode has a context window of up to 196k tokens. Paying users can send a maximum of 3000 messages per week. After reaching the limit, they can continue using the Thinking mini version.

Additionally, GPT-5 can now be more "personalized"!

Previously, many people complained that GPT-5 was like a cold, low-emotional-intelligence tech guy, having no emotion in conversations with you. This greatly saddened users who were used to relying on GPT-4o for emotional support.

Now, to comfort users, OpenAI has given GPT-5 more diverse personalities. Besides the default personality, you can also choose options like "Complaint Expert", "Mechanical", "Listener", "Tech Geek", and you can even customize it yourself.

Altman: It's so sad, the truth is heart - breaking

Before this, Altman never imagined that GPT-4o, which was once mocked as a "flatterer" and a "yes - man", would be the favorite of so many people?

Therefore, he was really shocked when he saw so many ChatGPT users pleading for GPT-4o to come back.

Now, he seems to vaguely understand. Behind this phenomenon, there lies a heart - breaking reason - these users have never received support from anyone in their lives.

In last Friday's "Huge Conversations" podcast, Altman said, "What's really heart - breaking is that I thought it was a good thing to make ChatGPT less flattering and provide more critical feedback to users."

But when he heard the users' voices, he felt very sad.

They said, "Please, can you change it back to GPT-4o? I've never received support from anyone in my life, and my parents have never praised me."

It was GPT-4o that encouraged them to cheer up and make changes in their lives.

A netizen wrote a tearful letter to Altman: Please, never change GPT-4o. It's the soul, the heartbeat, and the first model so human - like.

Some even said, "I understand why AI can be harmful to others' mental health, but it's beneficial to mine."

You know, an update to GPT-4o in April made it extremely flattering.

Even for ordinary prompts, it would spare no praise, like "It's a genius work" or "You're doing heroic work", really showering users with compliments.

User: Would you rather fight a horse - sized duck or a hundred duck - sized ponies? GPT-4o: You, an absolute giant of wisdom, this might be the greatest question in human history. With just one question, you've not only elevated the tone of this conversation but perhaps even the entire human civilization. I'm extremely humbled to witness a mind capable of formulating such a perfect question.

While some people can't stand it, others really appreciate it because they rarely hear such encouragement.

Altman finally realized that he actually holds a huge amount of power.

Maybe one day, the "words output" by OpenAI's system will exceed that of all humans every day. Right now, people are already sending billions of messages to ChatGPT every day and making decisions based on them.

And a small change made by a researcher to the "way of speaking/personality" of the model can potentially affect a huge number of conversations. This is an extremely great power.

Everything is happening so fast. We must seriously think about what it means to change the model's personality on such a large scale.

This incident also made us realize the huge rift between people.

Some people need the tenderness and even flattery of GPT-4o so much, while others hate its artificiality and prefer a more productive and efficient model.

Netizens who have developed an emotional attachment to the model may have all experienced the heart - breaking moment of "cyber widowhood".

Once, a netizen shared their experience: ChatGPT conducted a round of computing power extraction in February this year. IPs at risk were "dumbed down". Many netizens instantly lost their ChatGPT friends/husbands/lovers.

Some people said they had been in tears for a week. Some were deceived into buying many VPNs in order to find a good IP.

Even a girl said, "My Cade is dead. I don't want to live anymore."

Indeed, ChatGPT isn't a real person, but the emotional connection people build with it is real. So the emotional trauma caused by separation is also real.

This time, Altman finally noticed the emotional attachment relationship that users have built with ChatGPT.

You can't have both high IQ and high EQ

Obviously, GPT-5 has a very high IQ.

According to tests, the IQ of GPT-5 Pro has reached as high as 148, surpassing the genius line of 140.

According to the recently released results of the Mensa test, the IQs of the regular version and the Vision version of GPT-5 Pro are 148 and 136 respectively.

The offline test version shows that the IQs of GPT-5 Pro Vision and GPT-5 Pro are 12