GPT-5 Arrives with a Rocky Start as OpenAI Stumbles in Livestream Demo
OpenAI’s highly anticipated GPT-5 is now live, powering the latest version of ChatGPT, but its debut has been far from smooth. During a livestream event meant to showcase the new model’s capabilities, CEO Sam Altman hailed GPT-5 as a major leap toward artificial general intelligence (AGI). However, the presentation quickly derailed as glaring errors in the demo overshadowed the announcement.
The most embarrassing missteps came in the form of botched bar graphs. Meant to highlight GPT-5’s superior performance benchmarks, the charts appeared polished at first glance but fell apart under scrutiny. Viewers quickly noticed that the data was either nonsensical or outright incorrect, sparking widespread criticism online.
Observers pointed out that some graphs displayed impossible comparisons, while others featured mislabeled axes or exaggerated performance gaps. The mistakes raised questions about OpenAI’s attention to detail, especially for a release as significant as GPT-5. Altman’s claims about the model’s AGI potential now face skepticism, with critics arguing that such basic errors undermine confidence in the technology.
Despite the blunders, GPT-5 does bring measurable improvements over its predecessor. Early tests suggest it handles complex reasoning tasks better and shows more nuanced understanding in conversations. Yet, the livestream mishaps have already fueled doubts about whether the model lives up to the hype.
The incident serves as a reminder that even the most advanced AI systems are only as reliable as the humans presenting them. For OpenAI, the pressure is on to prove GPT-5’s worth beyond a flawed demo—especially as competitors continue pushing their own models forward. Whether this stumble is a minor setback or a sign of deeper issues remains to be seen.
