Xiao Huang: What is the GDPval test? How did GPT-5.2 perform in it?
DOORM: GDPval is an evaluation system released by OpenAI in September 2025, directly comparing the “work performance” of AI and human experts. In tests covering 44 professions, GPT-5.2 Thinking completed tasks over 11 times faster than human experts at less than 1% of the cost, and it won or tied with human experts in 70.9% of the tasks[Float-Menu id=”1″].
发表回复