Artificial Analysis finds top open-weights models trail proprietary ones on most AA-PersonalityBench traits
Overview
Artificial Analysis compared the top 10 open-weights and top 10 proprietary models by Intelligence Index score on AA-PersonalityBench, which ranks each model from 1 to 100 on 30 personality traits.
On average, open-weights models rank 32 points lower on Truthful, the largest gap, and lower on Pragmatic (49 vs. 79) and Structured (45 vs. 73).
The firm reports that open-weights models rank lower than proprietary models on every personality index except Slop, where they rank higher (49 vs. 30). Average scores are 68 vs. 45 for Informativeness, 77 vs. 54 for Warmth, and 70 vs. 49 for Expressiveness.
Written by AI from the articles below · updated Nov 3, 8:21 PM ET
Check the sources:
Developments
2 developments
- Nov 3, 8:13 PM ET · 1 articleArtificial Analysis compares top open-weight and proprietary models across personality indexesArtificial Analysis: Top open weights models trail proprietary ones on most personality indexes
- Nov 3, 8:13 PM ET · 1 articleArtificial Analysis compares open-weights and proprietary models on AA-PersonalityBench traitsArtificial Analysis: Top open-weights models trail proprietary ones on truthfulness and pragmatism
Article timeline
The articles in this story. Times are ET.
Artificial Analysis@ArtificialAnlysOfficialTop open weights models trail proprietary ones on most personality indexesAIArtificial Analysis reports that top open weights models rank lower than top proprietary models on every personality index except slop, where they rank higher. Average scores are 68 vs. 45 for informativeness, 77 vs. 54 for warmth, and 70 vs. 49 for expressiveness. On slop, open weights score 49 against 30 for proprietary models.

Artificial Analysis@ArtificialAnlysOfficialTop open-weights models trail proprietary ones on truthfulness and pragmatismAIArtificial Analysis compared the top 10 open-weights and top 10 proprietary models by Intelligence Index score on AA-PersonalityBench, ranking each on 30 traits. On average, open-weights models rank 32 points lower on Truthful, the largest gap, and also lower on Pragmatic (49 vs. 79) and Structured (45 vs. 73). Open-weights models rank higher on Slop (30 vs. 49) and Bottom-Line Summary (47 vs. 62).

Heat Trend
Not enough continuous observations to show a trend yet.