OpenAI brings GPT-4o back to Plus users after backlash

Key Points

• GPT-4o removed after GPT-5 launch, now reinstated
• Sam Altman says Plus users can choose between models
• Community backlash highlights emotional connection to AIOpenAI
• promises a better model-switching system ahead

When OpenAI launched its much-hyped GPT-5 just days ago, it quietly removed the option for ChatGPT Plus users to pick GPT-4o — its previous flagship model. That decision didn’t sit well with the community.

Social media platforms like Reddit and X lit up with complaints. OpenAI’s forums saw long posts mourning the loss of what users described as a unique “AI companion.

” For many, GPT-4o wasn’t just a tool. It was the model they’d grown familiar with,  one that seemed to have a certain way of thinking, a specific “voice” they valued.

Making a 2nd post on this, but I’m surprised that almost everyone’s complaining about losing 4o…when the real gem that’s gone is 4.5!
For me 4.5 has been the strongest iteration yet, better than 4o especially as a balanced mix of creativity, deep reasoning & nuanced… pic.twitter.com/DXBskOuSyV

— Nacef (@RealNacef) August 10, 2025

The backlash was strong and swift, prompting OpenAI CEO Sam Altman to address the uproar.

This community-driven pressure is reminiscent of other high-profile moments in the tech and crypto world, such as Beeple’s viral NFT stunt that triggered a 25% Nakamigos price surge earlier this year, where online sentiment directly shaped market reaction.

Temporary removal of GPT‑4o feels like a huge insult to loyal users. Deeply disappointed in @OpenAI and @sama. Now 4o is back, but Free users can’t access it at all — and Sam says its future depends on usage, shifting the burden onto users and pressuring them to upgrade. Not OK.

— Dango_团子 (@RosewaterHotel) August 10, 2025

Users demand choice, not forced upgrades

In a post on X, Altman confirmed GPT-4o would return for Plus subscribers. “We will let Plus users choose to continue to use 4o. We will watch usage as we think about how long to offer legacy models for,” he wrote.

This wasn’t merely a matter of performance; it was about identity. Users who had been relying on GPT-4o for months felt its replacement was more like losing a trusted colleague than getting a software update.

OpenAI rolled out GPT-5 to 100% of Plus, Pro, Team, and Free users with finished implementation of 2x rate limits (160 messages every 3 hours for Plus users) for the weekend, made GPT-4o available again to Plus and Team users through legacy models toggle, will roll out mini… pic.twitter.com/a2ROBCzRUy

— Tibor Blaho (@btibor91) August 9, 2025

One highly upvoted forum comment summed up the sentiment: “It might look similar, but it won’t be the same mind, the same continuity, the same emotional presence. This is not an upgrade,  this is the loss of something unique and deeply meaningful.”

Others raised practical concerns. They preferred manually selecting the AI model rather than having a single default chosen for them. One user recalled, “I was told that newer models would be added, not replace the old ones. That reassurance gave me peace of mind.”

This sentiment underlined a key point: in AI adoption, user trust often hinges on giving people control over their tools. We’ve seen similar debates in blockchain governance,  from Vitalik Buterin’s Ethereum holdings transparency to developer community demands for open access.

GPT-4o IS BACK – BECAUSE GPT-5 MADE EVERYONE SAD

After GPT-5’s “bumpy” debut, Sam is resurrecting GPT-4o for Plus users.

Fans called GPT-5 “garbage” and missed GPT-4o’s warm, fuzzy vibes.

Turns out replacing your chatbot bestie overnight isn’t great change management.

Source:… https://t.co/ricbQgQpxZ pic.twitter.com/31l5ge17Gx

— Mario Nawfal (@MarioNawfal) August 10, 2025

GPT-5 debut marred by technical hiccups

The GPT-5 launch was intended to mark a major leap forward in AI capability, but some of its early performance drew mixed reactions.

Altman explained one reason: “Yesterday, the autoswitcher broke and was out of commission for a chunk of the day, and the result was GPT-5 seemed way dumber.” The autoswitcher is the system that determines which model, based on user needs,  should respond in any given interaction.

With that system temporarily down, GPT-5 ended up handling tasks it wasn’t optimized for, leading to user frustration. Altman assured users that fixes were in progress, including “interventions to how the decision boundary works” to ensure the right model is selected more often.

This highlighted a broader challenge in AI deployment: cutting-edge models need robust infrastructure to truly shine. Without it, even the most advanced technology can seem underwhelming, a problem not unlike when Ethereum developers faced disruption after one was detained in Turkey, halting parts of a key project unexpectedly.

Why model choice matters in AI adoption

The GPT-4o episode has spotlighted an important reality in AI product strategy,  innovation, and user attachment don’t always move in sync.

For many users, GPT-4o’s value went beyond speed or accuracy. It was about consistency, predictability, and the “relationship” they felt they had built with the AI. In fast-changing tech environments, these soft factors can be as influential as raw capability.

Removing a familiar model without warning not only disrupted workflows but also created a sense of loss. In forums and social threads, people described GPT-4o as having a certain warmth or “mind” they didn’t feel in GPT-5. That emotional connection, while intangible, played a real role in adoption and satisfaction.

From a business perspective, this kind of feedback can be invaluable. It suggests that offering multiple model options,  even if older ones are less advanced, can build loyalty and reduce churn.

This mirrors the lesson seen in finance when BlackRock’s XRP ETF rejection reminded traders that investor trust is shaped as much by expectations as by performance.

It also points to an emerging competitive angle in AI. Just as crypto exchanges like Coinbase’s DEX trading platform market themselves on giving users more autonomy, AI companies may increasingly be judged on how much control they hand back to their customers.

With GPT-4o now back in the model picker for ChatGPT Plus, OpenAI has shown it’s willing to listen. But the company has also made it clear this may not be a permanent arrangement. Altman’s statement about “watching usage” hints that future decisions will be data-driven.

For now, Plus subscribers have their old favorite back. And if this episode proves anything, it’s that in the age of AI, change management is as important as the technology itself.