After a disastrous 72 hours that saw its most loyal users in open revolt, OpenAI is making a major U-turn.
In a series of posts on X (formerly Twitter) Sunday, CEO Sam Altman announced that the company is bringing back its beloved older AI models, including GPT-4o, and dramatically increasing usage limits for paying subscribers, a clear peace offering to a furious customer base.
The move comes just days after the botched rollout of GPT-5, the company’s latest and most powerful model. The launch, which should have been a triumph, instead sparked a firestorm.
the percentage of users using reasoning models each day is significantly increasing; for example, for free users we went from <1% to 7%, and for plus users from 7% to 24%.
i expect use of reasoning to greatly increase over time, so rate limit increases are important.
— Sam Altman (@sama) August 10, 2025
A Botched Launch and a User Mutiny
On August 7, OpenAI launched GPT-5, presenting it as a “unified system” that would automatically route user queries to the best model for the job. In doing so, it removed the menu that allowed users to choose between older, trusted models like GPT-4o, which was launched in March 2023.
For customers paying for subscriptions like ChatGPT Plus ($20/month), the sudden change felt like a betrayal. They had built their professional and creative workflows around a toolkit of different models, each with its own strengths: one for creativity, another for pure logic, another for deep research. Forcing everyone onto a single new model broke those workflows and removed their ability to cross-reference answers to check for errors or hallucinations. The backlash was immediate and fierce, resulting in a cascade of subscription cancellations and online petitions.
Altman’s Public Reversal
As the outrage grew over the weekend, Sam Altman took to X to do damage control, acknowledging the company had misjudged the situation and announcing a series of concessions.
The biggest news came in a direct reply to a user asking the question on everyone’s mind: “And bringing back 4o?”
“It’s back! go to settings and pick ‘show legacy models’,” Altman responded, confirming the return of the fan-favorite model. While GPT-5 remains the default, users can now opt back into the older versions.
it’s back! go to settings and pick “show legacy models”
— Sam Altman (@sama) August 10, 2025
A user then jokingly asked if they would be “marked as weird” for using the legacy models. “Not marked as weird in either case!” Altman replied.
To further appease paying customers, Altman announced a massive increase in usage limits for GPT-5’s most powerful features. “Today we are significantly increasing rate limits for reasoning for chatgpt plus users, and all model-class limits will shortly be higher than they were before gpt-5,” he posted.
When a user asked for specifics on the new limits, Altman revealed the new cap: “Trying 3000 per week now!” That’s a huge increase for a $20 per month subscription and a clear incentive for angry users to stick around.
trying 3000 per week now!
— Sam Altman (@sama) August 10, 2025
A Promise of Transparency
Finally, Altman promised more transparency. He announced an upcoming UI change to show users which model is actively responding to their queries and promised a detailed blog post this week explaining the company’s “thinking on how we are going to make capacity tradeoffs.”
He also shared data showing the immense popularity of the new reasoning models, revealing that daily usage among Plus users has jumped from 7% to 24%, subtly justifying the company’s initial focus on the new tech, even if the rollout was a disaster. “Reasoning” is the AI’s ability to “think” step-by-step to solve a complex problem.
In the end, OpenAI seems to have gotten the message loud and clear. This was a rare and powerful display of user revolt, forcing one of the most powerful companies in tech to listen and backtrack. The compromise—keeping GPT-5 as the default while giving users back their choice—is a direct recognition of the power of a loyal community that refuses to be ignored.