OpenAI: GPT-5 performs on par with humans in many professional fields.

On September 25th, OpenAI released a new benchmark test aimed at evaluating the performance differences between its artificial intelligence models and human professionals in various industries and occupations. This test, called GDPval, is a preliminary attempt by the company to understand how close its systems are to human performance in "high economic value work" - a key component of OpenAI's mission to develop general artificial intelligence. OpenAI stated that research found their GPT-5 model and Anthropic's Claude Opus 4.1 model are "approaching industry expert level in terms of work quality."