OpenAI Releases MedChemBench Medicinal Chemistry Benchmark
OpenAI has published MedChemBench, a medicinal chemistry benchmark built from 117 papers and patents spanning 761 graded molecular structures, with prompts cut off at real decision points and later reported results withheld. The strongest model, GPT-6.2 Astra, scores about 84% overall and is near-perfect at reading chemistry, yet is correct only about half the time on the decision that actually matters, exposing a clear gap in real-world drug design judgment.
Related Entities
Full Report
Rohan Paul (@rohanpaul_ai) · 2026-11-20 21
OpenAI has now published MedChemBench, its medicinal chemistry benchmark.
Its problems come from 117 papers and patents, span 761 graded molecular structures, and withhold the results researchers reported later.
Each prompt stops at a real decision point, supplying the biological target, the compounds made so far and selected assay results.
the best model GPT-6.2 Astra gets about 84% overall, is near perfect at reading the chemistry, and is right about half the time on the decision that actually matters.
