ChatGPT to Keep Old Models After GPT-5 Backlash
▼ Summary
– OpenAI will no longer retire old models without advance notice after backlash over replacing GPT-4o with GPT-5.
– OpenAI acknowledged user attachment to GPT-4o and is working to incorporate its “warmth” into GPT-5 while reducing perceived annoyances.
– The initial removal of GPT-4o aimed to simplify model choices for ChatGPT’s 700 million weekly users, prioritizing ease of use over customization.
– GPT-4o has been reinstated as an opt-in model for paying users following strong protests from vocal power users.
– OpenAI plans to provide more predictability for major changes, mirroring practices already in place for enterprise users.
OpenAI has reversed course on its decision to retire older ChatGPT models following significant user backlash over the abrupt removal of GPT-4o. The company now plans to maintain previous versions alongside new releases, acknowledging the unexpected emotional connection users developed with specific AI personalities.
Nick Turley, OpenAI’s head of ChatGPT, admitted the initial move to sunset GPT-4o was a misstep. “We underestimated how attached people became to certain model behaviors,” he explained. Many users likened the sudden shift to GPT-5 as losing a trusted companion, with complaints flooding forums about the newer model’s terser, less engaging responses. In response, OpenAI reinstated GPT-4o as an opt-in option for paying subscribers.
The original rationale for retiring older models centered on streamlining the user experience. With over 700 million weekly users, most rely on default settings rather than manually selecting AI versions. Turley emphasized that cost wasn’t the driving factor, simplicity was. “People want the product to adapt to their needs without requiring technical decisions,” he noted. Yet the outcry revealed a vocal minority of power users who deeply valued customization.
To address future transitions, OpenAI will implement clearer communication. Turley confirmed the company won’t phase out models without warning, mirroring policies already in place for enterprise clients. “Predictability matters at this scale,” he said, referencing plans to extend advance notices to all users.
Despite initial turbulence, GPT-5’s launch correlated with increased platform engagement overall. Turley highlighted the challenge of balancing feedback from advanced users against mainstream adopters experiencing AI reasoning capabilities for the first time. “We’re learning how to navigate these shifts without alienating either group,” he added.
The full interview with Turley will air later this week, offering deeper insights into OpenAI’s evolving approach to model management. For now, the reinstatement of GPT-4o signals a more user-centric strategy, one that prioritizes flexibility alongside innovation.
(Source: The Verge)
