OpenAI's GPT-5 Livestream Charts Reveal Inconsistencies
By
Jay Peters
Toasted to a respectable shade. No regrets, no crumbs left.
Summary
OpenAI's GPT-5 livestream showcased charts with inconsistencies, such as a misleading graph on 'deception evals across models.' The CEO acknowledged the error, calling it a 'screwup.' The article highlights discrepancies between the livestream and the official blog post.
Key quotes
· 2 pulledFor 'coding deception,' the chart shown onstage says GPT-5 with thinking apparently gets a 50.0 percent deception rate, but that’s compared to OpenAI’s smaller 47.4 percent o3 score which somehow has a larger bar.
CEO Sam Altman called one a 'screwup.'
You might also wanna read
OpenAI's GPT-5 Release Fails to Meet Expectations, Diminishing AI Hype
The article analyzes the recent decline in AI hype following OpenAI's disappointing release of GPT-5 on August 7th. The author argues that t
OpenAI's GPT-5 Release Falls Short Amid High Expectations
The article discusses OpenAI's release of GPT-5, which was highly anticipated but ultimately disappointing due to a poorly executed keynote.
OpenAI Researcher's GPT-5 Math Breakthrough Claim Retracted After Community Criticism
An OpenAI researcher claimed on X that GPT-5 had solved 10 previously unsolved Erdős mathematical problems and made progress on 11 more, but
GPT-5's Disappointing Release and the Fallout from a Troubling Research Paper
The article critiques the underwhelming release of GPT-5 by OpenAI, highlighting its delayed arrival and lackluster performance. It also men
GPT-5 Launch Raises Concerns Over Performance and Expectations
The article discusses concerns about GPT-5's launch, highlighting its failure to correctly answer a classic LLM trick despite being touted a
GPT-5's Disappointing Launch Exposes AI Industry Realities
The article critiques the underwhelming launch of OpenAI's GPT-5, highlighting its implications for the AI industry, including the gap betwe
