8 months after it was removed from the shelves, GPT-4o’s rescue operation is still going on

📅 2026-10-11

Abstract:

Every day. 14 months, 8 months, and we're still here. This is not stubbornness, delusion, or illness. A Keep4o supporter wrote this on X. What he is waiting for is

GPT-4o

who retired from ChatGPT in February this year . Nearly 8 months have passed and no one has said goodbye. Now, another date is in front of us:

October 23rd

.

OpenAI will deactivate the earliest API snapshot of GPT-4o. Its name says "2024-05-13", which is the day 4o made its first appearance.


Other versions of the API are still there, but the earliest version will no longer be available to the public.

In the past two days, "The campaign to save GPT-4o continues" has become popular on the X trend page again.


Let me tell you, 4o fans, you are so persistent...

This is already GPT-6, why bother saving it?

Some people are still looking forward to its return, and some people have made new requests:

If one day you no longer need it, can you at least keep its weight?

What exactly are you doing to save GPT-4o?

The latest round of action is a new plan posted on X by netizen BlueBeba on October 7, titled

"GPT-4o archived open weight"

.

He proposed that OpenAI should set an archiving policy for retired models. 4o Since it is no longer under development, release its weight according to the appropriate license.


The plan mentioned that the training data is not required, the training process is not required, and internal tools and post-training methods are not required. The only thing required is the trained model itself.

According to his metaphor, they just want to keep the cooked dishes and do not ask for the recipes to be made public. In this way, OpenAI does not have to provide services to it all the time.

Even at the end of writing, BlueBeba’s tone became anxious, blaming OpenAI:

You say it is outdated, so outdated that it can be retired. But you think it is valuable, so valuable that you want to lock it up forever.

He is a long-time supporter of Keep4o. If you carefully look at his social media, you will find that he used GPT-4o to conduct a real-time video Q&A two days ago.


In his opinion, 4o can catch the action in the screen, and can also follow the jokes and tone.

He wrote: "Running scores is cheap, but understanding people in real time is priceless."

This is just his impression of one attempt, but it goes a long way to explaining why Keep4o fans love this model so much.

So, how many people are still thinking about 4o?

On Change.org, there is a document called

"Please Keep GPT-4o Available on ChatGPT"

petition.

As of today, 23,930 people have signed it.


It started soliciting in April 2025, earlier than 4o was first withdrawn. The appeal is to continue 4o as an optional model in ChatGPT and API.

The last sentence of the petition is: "Thank you for giving us something worth cherishing."

In more than a year from the creation of this petition to now, 4o fans have tried various methods in the hope of keeping this model.

Set the time back to August 2025, the day GPT-5 went online, 4o was directly removed.

Users flooded into Ultraman’s Reddit Q&A, saying that the new model “has become dumber” and begging OpenAI to return 4o.

Ultraman quickly relaxed: Plus users can continue to choose 4o, "We will observe the usage and decide how long to keep the old model."


After a few days, 4o came back. This is also the only time the Keep4o movement has won.

At the end of September 2025, another user discovered that when talking about emotions or sensitive topics, they clearly chose 4o, but the answer was quietly changed to another security model.

So a new petition appeared again, and this time everyone’s appeal became: Stop using other models to pass it off.

Nick Turley, the person in charge of ChatGPT, publicly explained that this was a safety mechanism, but related actions have not stopped.


Until January this year, OpenAI announced that it would officially remove 4o from the shelves on February 13.

The reason given by OpenAI is: Most users have switched to GPT-5.2,

Only 0.1% of users still choose GPT-4o every day

.

The fans immediately went crazy. The old petition on Change.org was resurfaced, and the number of signatures increased to over 20,000.

Some people posted screenshots of unsubscribing, and some went to Ultraman’s live broadcast comment area to swipe the screen.


It was also during that period that the appeal for the first time retreated to "open source."

On February 2, a new petition was launched, hoping that OpenAI would release the weights and architecture of GPT-4o under a non-commercial research license, allowing the community to preserve and study it.

This petition also refutes the 0.1% figure: The system has been automatically switching users to GPT-5 for several months. If you want to use 4o, you have to manually switch back. Free users cannot choose at all.


However, this petition has only 472 signatures so far.

OpenAI has not changed its decision to remove it from the shelves. On February 13, 4o disappeared from ChatGPT.

On February 28, more than 30 people held a vigil outside the OpenAI headquarters in San Francisco.

They held slogans, brought hundreds of paper cranes, and posted real messages one by one about how 4o helped their daily lives.


One participant posted afterwards: "We made our demands, but when we left, we left nothing behind."

At the end of July this year, the opensource4o Alliance, led by an Austrian digital rights organization, issued a public statement requiring OpenAI to open the March 2025 version of 4o weights under the Apache 2.0 license.


Their reason is: In the GPT-4o system card in August 2024, network security, biological risks, and model autonomy are all rated as low risks, and the weights should be disclosed.

The statement also left a way out: even if you haven't decided whether to open it or not, at least make a public commitment first and don't delete it.

From "Give it back to us", to "Don't fake it", "Open source it", and then "Don't delete it", for more than a year, 4o fans have always remembered this model, but they want less and less.

There has been no response from OpenAI.

GPT-4o, the life of "Moonlight"

How charming is GPT-4o that can make a group of people never forget it?

The story must start from the first time it speaks.

On May 13, 2024, 4o made its first appearance, and it was like a movie.

o is omni, which means "all-powerful". It can directly understand and respond to text, pictures, and sounds.

During the demonstration, it can be interrupted at any time, will answer the conversation, and will laugh out loud when hearing the joke. If the researcher speaks breathlessly on purpose, it will remind the other person to slow down first.

Related tags

Related articles

Comments

0/500
Captcha (click to refresh)
No comments yet