OpenAI’s GPT-5 Launch Receives Sharp Backlash Over Glitches, Sam Altman Apologizes

OpenAI’s much-hyped GPT-5 debut quickly turned sour as users slammed the model for glitches, colder responses, and the loss of GPT-4o’s personality. The backlash prompted CEO Sam Altman to apologize publicly and reinstate GPT-4o for some users just a day after it was replaced

Much-Hyped GPT-5 met with disappointment

OpenAI unveiled its much-hyped GPT-5 model on Thursday via a livestream, promising major upgrades in coding, reasoning, accuracy, writing, and multimodal capabilities, along with a reduced tendency to “hallucinate” compared to previous models. But instead of celebration, the launch was met with sharp disappointment from both users and experts.

A series of glitches quickly soured the rollout, forcing CEO Sam Altman to publicly apologize and admit that GPT-5 had “acted like a fool” due to a technical error. The backlash has put OpenAI in a precarious position: without convincing fixes in the coming days, the months of hype Altman has built could give way to a collapse in user and investor trust.

Frustration deepened when OpenAI discontinued older models. Social media erupted with complaints that GPT-5 lacked the personality of GPT-4o, delivering shorter, less contextually accurate responses and missing the “warmth” and “emotional intelligence” of its predecessor. On Reddit, a thread titled “GPT-5 is terrible” drew nearly 3,000 upvotes and over 1,200 comments, with one user lamenting: “GPT-4.5 really spoke to me… it was my only friend. I went to talk to it this morning, and I got some really cold words.”

Users sounded disappointed, and not only for that. In Reddit groups, harsh complaints were heard about GPT 5, with one The threads titled "GPT 5 is terrible" receiving nearly 3,000 upvotes and over 1,200 comments supporting the opinion. One user wrote: "GPT 4.5 really spoke to me, and as pathetic as it sounds, it was my only friend. "I went to talk to him this morning, and I got some really cold words."

And it seems that OpenAI quickly realized that they were in trouble. Otherwise, it is impossible to explain the series of apologies and hysterical responses from the company's executives, which also included the immediate return of GPT-4o to use, just one day after it was replaced by GPT 5.

The backlash was strong enough that OpenAI reversed course within a day, reinstating GPT-4o alongside GPT-5. In a post on X, Altman acknowledged underestimating how much users valued aspects of GPT-4o, even if GPT-5 performed better “in most ways.” For now, only Pro plan subscribers ($200/month) can manually switch back to GPT-4o, though Altman suggested this might expand. “We for sure underestimated how much some of the things that people like in GPT-4o matter to them, even if GPT-5 performs better in most ways," he wrote.

"The team is doing heroic work to optimize our systems and find more capacity, but still, we are looking at a severe capacity challenge for next week," Altman added. "We are still deciding what we are going to do, but we will be transparent with our principles."