Top AI Models Get About a Quarter of Factual Questions Wrong Without Search

Data from Artificial Analysis' AA-Omniscience benchmark, covering 6,000 factual questions across 42 domains, shows top AI models get roughly a quarter of answers wrong without search, highlighting the need for verification of important facts.

2026-09-30 ~ 2026-09-30 · 2 related posts

• Even top AI models get about 1 in 4 facts wrong without search, Artificial Analysis data shows — randal_olson · 2026-09-30
• AA-Omniscience benchmark: 6,000 fact questions show top models still err ~25% without search — randal_olson · 2026-09-30