OpenAI walks back GPT-5 rollout after user backlash

OpenAI CEO Sam Altman has admitted the company "totally screwed up" the launch of its GPT-5 model, after abruptly replacing all previous ChatGPT versions on release day and triggering a wave of complaints and subscription cancellation threats.

GPT-5, rolled out on August 7, was pitched as a major leap in coding, reasoning, factual accuracy, health, and multimodal abilities. 

But long-time users complained it gave shorter answers and lacked the emotional depth of GPT-4o. Within days, OpenAI reinstated GPT-4o for Plus subscribers and increased usage limits for GPT-5's Standard and Thinking modes.

Speaking to reporters Thursday, Altman said the company underestimated the attachment users had to older models: "We've learned a lesson about what it means to upgrade a product for hundreds of millions of people in one day."

According to OpenAI, GPT-5 offers stronger reasoning and coding performance, higher factual accuracy, broader health and multimodal integration that seamlessly combines text, images, and audio, and faster processing with adaptive thinking capabilities.

Despite the rocky start, OpenAI says ChatGPT API traffic doubled in the first 48 hours post-launch and app usage hit an all-time high. The company recently announced ChatGPT reached 700 million weekly users, up from 500 million in March.

OpenAI's previous major update, GPT-4o in May, added native image generation and fueled a viral Studio Ghibli–style art craze. GPT-5's challenge now: deliver its technical gains without alienating its most loyal base.