BehaviorBench - Cross-Dimension Personality: leaderboard
Metric: Mean absolute error of the predicted score on one Big Five dimension from the other four dimension scores (Big Five personality dimension scores of held-out survey subjects); lower is better. Source: arxiv.org. Saturation forecast: Around August 2027. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 | 6.7 |
| 2 | Gemini 3.1 Pro (Preview) | 6.8 |
| 3 | Gemini 3.1 Flash Lite (Preview) | 7.1 |
| 4 | Claude Sonnet 4.6 | 7.2 |
| 5 | GPT-4.1 | 7.2 |
| 6 | Claude Haiku 4.5 | 7.2 |
| 7 | GPT-5.4 (High) | 7.3 |
| 8 | GPT-5.4 Mini (High) | 7.4 |
| 9 | DeepSeek V3.2 | 7.5 |
| 10 | Qwen 3 4B | 7.6 |
| 11 | Llama 3.3 70B Instruct | 8.3 |
Interactive version: theaggregate.ai/benchmark?slug=behaviorbench-cross-dimension-personality · How It Works · Data refreshed daily, snapshot 2026-09-29.