OpenAI's newly released GPT-5 Turbo has demonstrated near-human performance on a battery of graduate-level science examinations, scoring in the 91st percentile on physics, 94th on chemistry, and 89th on biology GRE subject tests. The model also passed the United States Medical Licensing Examination with a score equivalent to the top 15% of human test-takers.
Researchers note that GPT-5 Turbo's improvements stem largely from enhanced reasoning capabilities rather than simply memorizing more data. The model shows marked improvement in multi-step problem-solving and the ability to integrate concepts across scientific disciplines.
