OpenAI is scrambling to fix GPT-5 after a rough launch on Thursday. Users say GPT-5 feels “dumber” than the older GPT-4o. The problem? A real-time router that decides which model answers your query wasn’t working right.
During a Reddit AMA, CEO Sam Altman explained the autoswitcher went down for a chunk of the day, causing the dumbed-down experience. He promised fixes starting immediately.
Altman also addressed the demand to bring back GPT-4o for Plus subscribers. 
“We are looking into letting Plus users to continue to use 4o. We are trying to gather more data on the tradeoffs,” Altman said.
“We are going to double rate limits for Plus users as we finish rollout.”
Users should get more prompts to test the new GPT-5 without hitting limits.
The AMA also touched on the notorious “chart crime” from the GPT-5 launch event—a wildly misleading chart with a low score but a huge bar. Altman didn’t comment during the AMA but called it a “mega chart screwup” on X.
Some early reviewers, like Simon Willison, pointed out GPT-5 struggles with turning data into tables—a clear weak spot.
Altman wrapped the AMA with this:
“We will continue to work to get things stable and will keep listening to feedback.”
OpenAI is clearly reacting fast. But the GPT-5 rollout still has some growing pains.
