Top Clinical AI Models Drop to 65% on Harder Medical Benchmark

The Clinical System One v0.2 benchmark, covering 43 models and 993 clinical decisions, shows top clinical AI models scoring 96-97.5% on the original test set but only up to 65% on a harder 324-question subset.

2026-10-27 ~ 2026-10-27 · 2 related posts

• Clinical System One v0.2: top clinical AI drops from 97% to 65% on harder cases — MaziyarPanahi · 2026-10-27
• Clinical AI benchmark: top models hit 96-97.5% on original test but only 65% on 324 harder cases — baseten · 2026-10-27