During a Reddit AMA on Friday, OpenAI CEO Sam Altman and key members of the GPT-5 team fielded a wave of questions about the new model — and calls to bring back the previous GPT-4o.
A central topic of frustration was GPT-5’s performance compared to 4o. Altman explained that a key new feature, a real-time router designed to choose the best model for each prompt, wasn’t functioning correctly when GPT-5 rolled out on Thursday.
The Glitch Behind GPT-5’s “Dumber” Performance
Altman acknowledged the issue, saying:
“GPT-5 will seem smarter starting today. Yesterday, we had a sev and the autoswitcher was out of commission for a chunk of the day, and the result was GPT-5 seemed way dumber.”
He added that adjustments are being made to improve the model selection process and promised more transparency on which model answers each query.
Plus subscribers might also see GPT-4o return. Altman confirmed the team is “looking into letting Plus users continue to use 4o” while gathering data on tradeoffs. Additionally, rate limits for Plus users will be doubled once the rollout is complete.
The Infamous “Chart Crime”
Another hot topic was the now-infamous “chart crime” from the GPT-5 launch presentation — a visual that showed a lower benchmark score with a much taller bar. While Altman didn’t address it during the AMA, he had already called it a “mega chart screwup” on X.
Although the charts in OpenAI’s official blog post were correct, the flawed presentation version became a viral joke, with some suggesting GPT itself had made the chart.
GPT-5 reviewer Simon Willison, who generally praised the model, also noted that converting data into a table was “a good example of a GPT-5 failure.”
Altman closed the AMA with a reassurance:
“We will continue to work to get things stable and will keep listening to feedback.”
The conversation made it clear that OpenAI is actively adjusting GPT-5’s performance while also weighing user requests for more model options and higher usage limits.
Author’s Opinion
The GPT-5 AMA reinforced a pattern: OpenAI often releases major updates before ironing out key technical issues. While early access generates excitement, it also risks alienating loyal users when glitches overshadow improvements. A short delay for extra testing could have avoided the “dumber” day entirely — and perhaps kept the focus on what GPT-5 does better, instead of its launch stumbles.
Featured image credit: Solen Feyissa via Unsplash
For more stories like it, click the +Follow button at the top of this page to follow us.