AA-PersonalityBench measures 30 language-model personality traits and 21 communication characteristics

ArtificialAnlys announced AA-PersonalityBench to compare how language models communicate, even when their answers are equally correct. It says LLM judges compare models’ answers to 900 questions across 30 subjective traits, which feed into eight personality indexes. The benchmark also calculates 21 communication characteristics, including length and punctuation. In its results, Qwen3.8 27B averages about 1,030 words per response, compared with GPT-6 Luna’s 270.