OpenAI restores GPT-4o after users complain that new GPT-5 feels 'dumber'

OpenAI has restored its GPT-4o model after user backlash against the newly launched GPT-5, which many described as "dumber." CEO Sam Altman acknowledged the issues during a Reddit session, attributing some problems to a malfunctioning real-time router intended to optimize responses.

He stated, “GPT-5 will seem smarter starting today,” emphasizing ongoing adjustments to improve performance. The launch was further complicated by a misleading benchmark chart that led to online ridicule. Altman referred to it as a “mega chart screwup.” He assures that feedback will drive improvements and hopes to stabilize the product shortly.