BehaviorBench - Cross-Dimension Personality (Distribution): leaderboard
BehaviorBench (University of Michigan, 2026) evaluates foundation models on behavioral science tasks at the individual and population levels, using economic-game records, Big Five personality surveys and research literature. This board is the Wasserstein distance between predicted and observed Big Five dimension score distributions when predicting one dimension from the other four; lower is better.
Metric: Wasserstein distance between predicted and observed distributions of one Big Five dimension score predicted from the other four (Big Five personality dimension scores of held-out survey subjects); lower is better. Source: arxiv.org. Saturation forecast: Around July 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 3.6 | #94 |
| 2 | GPT-4.1 | 3.93 | #495 |
| 3 | Claude Sonnet 4.6 | 4.37 | #161 |
| 4 | GPT-5.4 (High) | 4.4 | #123 (GPT-5.4) |
| 5 | Claude Opus 4.6 | 4.49 | #102 |
| 6 | Claude Haiku 4.5 | 4.59 | #568 |
| 7 | DeepSeek V3.2 | 4.61 | #406 |
| 8 | Gemini 3.1 Flash Lite (Preview) | 4.98 | #418 |
| 9 | GPT-5.4 Mini (High) | 5.02 | #389 (GPT-5.4 Mini) |
| 10 | Llama 3.3 70B Instruct | 5.19 | #1357 |
| 11 | Qwen 3 4B | 5.21 | #2281 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=behaviorbench-cross-dimension-personality-distribution · How It Works · Data refreshed daily, snapshot 2026-11-13.