A recent assessment by OpenAI reveals that AI is swiftly closing the gap with human professionals in performing tasks that hold economic value. As per reports, on September 25 (local time), OpenAI unveiled a new evaluation tool named GDPval-v0. This tool is specifically crafted to gauge the performance of AI models in accomplishing "real-world work deliverables," such as legal documents, engineering blueprints, and nursing plans. GDPval draws on the nine industries that make the largest contributions to the U.S. GDP, encompassing sectors like healthcare, finance, manufacturing, and government. It covers 44 occupations, spanning from software engineers to nurses and journalists. For the initial version, GDPval-v0, OpenAI enlisted the expertise of senior professionals. These experts were tasked with comparing AI-generated reports with those crafted by other professionals and selecting the superior one. The findings demonstrated that in 40.6% of cases, GPT-5-high (a high-computing-power variant of GPT-5) was rated as either superior to or on par with industry experts. Meanwhile, Anthropic's Claude Opus 4.1 model was deemed not inferior to industry experts in 49% of the tasks. OpenAI clarified that although AI models are approaching the performance level of human experts, GDPval, at present, only encompasses a limited fraction of the tasks involved in people's actual work. It excludes interactive processes and practical operations, indicating that AI is still far from replacing human jobs. Nevertheless, the pace of evolution of AI models is truly remarkable. Their performance has improved by more than two-fold, with the win rate of GPT-4o rising from 13.7% in spring 2024 to 40.6% for GPT-5 in summer 2025. OpenAI is of the opinion that the progress made by GDPval is substantial, as it can assist professionals in saving time and allowing them to concentrate on more valuable work.