The GDPval study was created to measure how well AI models perform real-world, economically valuable tasks across key U.S. industries. It shows that frontier AI systems are steadily improving, in some cases approaching the quality of experienced professionals, especially when paired with human oversight. The findings suggest AI can save time and costs while improving accuracy, though challenges remain with instruction-following, formatting, and highly specialized work .
Source: https://openai.com/index/gdpval/