Sam Altman on GPT-5 Rollout Challenges and User Concerns
When OpenAI launched GPT-5, the AI community was eager to test its promised speed, intelligence, and new decision-routing feature. However, the rollout did not go as smoothly as expected. Many early users reported that GPT-5 seemed less capable compared to its predecessor, GPT-4o. In a candid Reddit AMA, OpenAI CEO Sam Altman addressed these concerns head-on, explaining what went wrong and outlining the steps being taken to improve the experience. Within the first 24 hours of release, a technical issue with the real-time router caused GPT-5 to underperform, leading to what Altman admitted felt like a “dumber” AI. The router, designed to select the best model for each query—either for speed or deeper reasoning—wasn’t functioning correctly during part of the rollout.
Image Credits:Justin Sullivan / Getty Images
Why GPT-5 Initially Underperformed
The core reason behind GPT-5’s initial performance issues lay in its autoswitching system. This new feature was built to analyze user prompts and route them to the most suitable model in real time. However, due to a server-level incident, this feature was down for a significant portion of launch day. As a result, GPT-5 often defaulted to faster but less complex reasoning paths, giving the impression that it was less intelligent than GPT-4o. Altman reassured users that fixes had already been deployed and improvements to the decision boundary were underway. This means users should now experience more accurate model selection, resulting in smarter and more relevant responses. Importantly, OpenAI plans to make the system more transparent so users will know which model is responding to their queries.
Community Feedback and the Possible Return of GPT-4o
One of the most discussed topics during the AMA was the request to bring GPT-4o back, especially from Plus subscribers who preferred its performance for certain tasks. While GPT-5 offers enhanced capabilities, some users still find specific strengths in 4o, particularly in creative or nuanced conversations. In response, Altman confirmed that the company is exploring ways to allow Plus users access to 4o alongside GPT-5. This move would give subscribers more flexibility, enabling them to choose the best model for their needs. He emphasized that OpenAI is gathering data on trade-offs before making a final decision, signaling that user feedback will play a significant role in shaping the outcome.
What’s Next for GPT-5 and Plus Users
Beyond fixing the routing issue, OpenAI is implementing changes to enhance the GPT-5 experience for paying customers. Altman announced that Plus users will see their rate limits doubled, allowing more opportunities to explore and adapt to the new model without restrictions. This increased capacity is designed to help users integrate GPT-5 into their workflows, test its reasoning abilities, and better understand how it compares to 4o. By addressing technical glitches quickly, listening to community input, and offering potential access to multiple models, OpenAI aims to ensure GPT-5 meets both performance expectations and user trust. While the rollout may have been “bumpy,” the company’s transparent communication suggests that the future of its AI models is being shaped in partnership with the very people who use them.
إرسال تعليق