OpenAI Reverses Course: Beloved GPT-4o Returns to ChatGPT After ‘Bumpy’ GPT-5 Rollout | User Backlash & Performance Concerns Mount
Key Takeaways
- OpenAI has swiftly reinstated GPT-4o as an option for paid ChatGPT users following widespread user demand.
- The initial rollout of GPT-5 was met with significant user dismay and criticism, with many mourning the replacement of models like GPT-4o and o3.
- GPT-5’s debut was marred by a “bumpy” experience and reported performance regressions, including a notable failure on a basic algebra problem.
Main Developments
The AI world witnessed a swift and unprecedented turn of events today as OpenAI, after just a single day, announced the return of its popular GPT-4o model to ChatGPT for paid users. The rapid reversal comes amidst a tumultuous and widely criticized rollout of the highly anticipated GPT-5, which was initially intended to completely replace all previous models on the platform.
The drama unfolded quickly, with a chorus of ChatGPT users voicing their dismay and disappointment almost immediately following the transition to GPT-5. Many took to social media and forums, including a particularly “spicy” Reddit AMA, to mourn the unexpected removal of the “workhorse” GPT-4o, and even other beloved models like ‘o3’, from the user interface. The outcry was so pronounced that OpenAI CEO Sam Altman himself took to X (formerly Twitter) to confirm the company’s decision: “We will let Plus users choose to continue to use 4o.” This swift capitulation underscores the immense pressure and user backlash that accompanied the GPT-5 debut.
The rollout has been widely described as “bumpy,” with multiple reports citing a less-than-smooth user experience. VentureBeat AI highlighted the general user dismay, noting that while the enterprise API for older models remained available, the public-facing ChatGPT saw a complete overhaul. More concerning, perhaps, were reports of GPT-5’s immediate performance. Despite being touted as a significant leap forward, the new model apparently stumbled on tasks that its predecessors might have handled with ease. One specific example cited by VentureBeat AI involved GPT-5 failing a simple algebra problem (5.9 = x + 5.11) that an elementary schooler could likely solve, raising immediate questions about its core capabilities and whether it truly represents a “step forward” as OpenAI intended.
Sam Altman, addressing the situation directly, acknowledged the “bumpy” nature of the GPT-5 rollout, a stark admission that the company’s latest flagship model didn’t live up to its internal expectations or, more crucially, its users’ immediate needs. This incident places considerable pressure on OpenAI to not only demonstrate that GPT-5 isn’t just an incremental update, but a genuinely superior advancement in AI, especially given the public’s heightened expectations and the company’s leading position in the AI race. The return of GPT-4o, therefore, serves as a significant concession, allowing users to revert to a familiar and reliable option while OpenAI presumably works to iron out the kinks and prove the true value proposition of GPT-5. This episode highlights the delicate balance between innovation, user experience, and managing public perception in the fast-evolving AI landscape.
Analyst’s View
This rapid reversal by OpenAI is a powerful testament to both the vocal influence of a dedicated user base and the inherent challenges of deploying next-generation AI models at scale. It suggests that while the race for advanced AI capabilities is relentless, the imperative for robust testing, seamless user experience, and genuine performance improvements cannot be sacrificed. OpenAI, perhaps under pressure to showcase GPT-5, appears to have underestimated the reliance users had on established models like GPT-4o and oversold the immediate stability or superiority of its successor. Moving forward, the industry should watch how OpenAI regains user trust and whether GPT-5 can truly justify its existence as more than just a marketing statement. This incident sets a precedent: user sentiment can directly dictate the availability of key AI tools, signaling that the “AI race” must also prioritize practical utility and reliability.
Source Material
- ChatGPT is bringing back 4o as an option because people missed it (The Verge AI)
- ChatGPT users dismayed as OpenAI pulls popular models GPT-4o, o3 and more — enterprise API remains (for now) (VentureBeat AI)
- OpenAI returns old models to ChatGPT as Sam Altman admits ‘bumpy’ GPT-5 rollout (VentureBeat AI)
- Sam Altman addresses ‘bumpy’ GPT-5 rollout, bringing 4o back, and the ‘chart crime’ (TechCrunch AI)
- OpenAI’s GPT-5 rollout is not going smoothly (VentureBeat AI)