OpenAI · Medicinal chemistry benchmark

MedChemBench

OpenAI has introduced MedChemBench, a benchmark of 99 problems drawn from 117 published papers and patents, covering 761 molecular structures. It tests whether models can reconstruct molecules, explain experimental evidence, rank compounds and predict properties. At the xhigh reasoning setting, GPT-6.2 Astra scored 84.5% of available credit, against 58.2% for GPT-5.6 Sol, and OpenAI says the results do not show whether models can discover new drugs. A preprint accompanies the release.

Open the source ↗

Similar Research & Safety releases

More from OpenAI

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.