Artificial Analysis Launches AA-PersonalityBench: 30 Traits, 8 Indexes, 53 Models

ArtificialAnlys · x · 2026-11-19

Artificial Analysis released AA-PersonalityBench, measuring how models communicate rather than whether they are correct: 30 personality traits judged pairwise by LLM judges and scored via Bradley-Terry Elo over 900 prompts, rolled up into 8 indexes (Expressiveness, Organization, Informativeness, Transparency, Warmth, Conviction, Imagination, Slop), plus 21 directly measured communication characteristics. Early finding: higher-effort variants score much higher on Evidence-Based (GPT-6.2 Astra 12→54, Sol 16→68) and Disclaims Assumptions (Astra 20→52, Sol 26→77).

Original post →

More from Models

Models channel →