During a Reddit Q&A session on Friday, OpenAI CEO Sam Altman and the GPT-5 team faced questions about the new model and requests for the return of GPT-4o. Attendees also inquired about the presentation’s most amusing blunder, the “chart crime.”
One new feature of GPT-5 is a real-time router that selects the appropriate model for a prompt, choosing between quick responses or more thoughtful answers. However, users in the AMA expressed dissatisfaction, noting that GPT-5 wasn’t performing as well as 4o. Altman explained that the issue stemmed from a malfunctioning router during Thursday’s rollout.
Altman assured participants that GPT-5 would appear smarter, as the issue had been addressed. Additionally, adjustments to the decision boundary will be made to enhance model selection transparency. Due to users’ strong demand, Altman promised to consider allowing Plus subscribers to continue using 4o and to gather data on the tradeoffs involved.
Altman also vowed to double rate limits for Plus users during the rollout, providing more opportunities to explore and adapt to the new model.
The Q&A touched on the now-infamous incorrect chart from the live presentation, which became a subject of “chart crime” jokes. Altman, who didn’t address the chart issue during the AMA, labeled it a “mega chart screwup” in a separate online statement. The accurate charts were found in the published blog.
Despite the blunder, Altman promised to fix the issues that concerned users. He concluded the AMA with a commitment to stabilize the platform and remain open to feedback.