Grok 4.7 cuts hallucination rate to 29% from 34% on AA-Omniscience

ArtificialAnlys · x · 2026-09-22

Artificial Analysis published the full Intelligence Index breakdown for Grok 4.7 (xhigh) alongside Grok 4.6 (high) and other leading models. On AA-Omniscience, hallucination rate dropped from 34% to 29%, accuracy held roughly steady (47% vs 48%), and the overall index improved from 30 to 32.

Related event: Grok 4.7 Scores 46 on Intelligence Index, Enters Top Four Labs at Double Token Cost(7 posts)→

Original post →

More from Models

Models channel →