OpenAI CEO Sam Altman and the brains behind GPT-5 recently jumped into the deep end of a Reddit AMA, and it didn’t take long for the community to fire off tough questions — from model performance to one rather infamous “chart crime.”
The buzz? GPT-5 has just launched with some flashy upgrades, including a real-time router that decides whether to give you a lightning-fast answer or take its sweet time “thinking” for better accuracy. Sounds great in theory… except, on launch day, it didn’t exactly work as intended.
Many Redditors claimed GPT-5 felt “dumber” than its predecessor, GPT-4o. Altman explained the culprit: the routing system was down for part of the day, meaning GPT-5 wasn’t switching between models as designed. His reassurance? “GPT-5 will seem smarter starting today,” adding that they’re tweaking the decision-making so users get the right model more often — and they’ll make it clearer which model is actually responding to your query.
But nostalgia is strong in the AI world. Several users lobbied hard for GPT-4o’s return, especially for Plus subscribers. Altman didn’t shut the door on the idea, saying they’re actively exploring whether to keep 4o around for paying members. As a sweetener, he also promised to double the rate limits for Plus users during the rollout, giving everyone more room to experiment without hitting a wall.
And then, the “chart crime.” During the big GPT-5 reveal, a benchmark bar chart showed a much lower score… with a much taller bar. Cue the memes. Altman later admitted it was a “mega chart screwup,” though the blog post version had it right. The internet, of course, had its fun, with jokes about letting GPT handle corporate slide decks.
Despite the hiccups, Altman closed with a promise: OpenAI is listening, fixing, and fine-tuning. Translation? The GPT-5 ride might have started bumpy, but they’re not taking their hands off the wheel.
About the Author
