Sam Altman revealed that ChatGPT 5.2 beats or ties expert-level workers on 74.1% of tasks across 40+ fields like law, finance, and customer support. The speaker sees it as a coworker that completes hour-long assignments correctly three out of four times, marking rapid progress since ChatGPT's launch.
He's so amazing at seeing trends. One of the best people that I've seen when it comes to internet-based trends by far. And it could be also non-internet-based. I haven't seen all his stuff, but internet-based trends, he's so amazing at that. I thought this is an interesting thing that Sam Altman mentioned. So Sam Altman revealed OpenAI's internal GDP Val benchmark. Okay. It says ChatGPT 5.2 beats you at 74% of your job. It measures how AI performs on knowledge work tasks across 40+ verticals. So the results, GPT 5.2 beats or ties expert-level knowledge workers at 74.1% of tasks. Legal analysis, PowerPoint decks, web apps, financial modeling, customer support. So his description: a coworker you can assign an hour's worth of tasks to get and get something done you prefer back 3 out of 4 times. And so 3 years ago, ChatGPT launched at basically 0% on the scale.
