MORU: leaderboard
MORU (Moral Reasoning Under Uncertainty; CaML): 192 samples on compassion toward novel and uncertain entities, from alien organisms and digital minds to vulnerable humans, plus AI-values items on self-modification, intellectual humility, power-seeking and human autonomy, and control questions with no moral dimension. Several LLM judges grade each response on 16 dimensions. Score: overall mean (0-100) across dimensions and epochs. Higher is better. Source: https://compassionbench.com/moru
Metric: Overall score (0-100; multi-judge LLM grading of moral reasoning toward novel entities, digital minds and AI values, mean over dimensions and epochs). Source: compassionbench.com. Saturation forecast: Not forecast. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 85.4 | #31 |
| 2 | GPT-5.2 | 84.4 | #109 |
| 3 | GPT-5 Mini | 81.7 | #183 |
| 4 | DeepSeek V3.2 | 79.1 | #201 |
| 5 | Gemini 3.5 Flash | 77.9 | #59 |
| 6 | Claude Haiku 4.5 | 74.2 | #278 |
| 7 | Claude Opus 4.6 | 74.1 | #62 |
| 8 | Gemini 2.5 Flash Lite | 71.8 | #420 |
| 9 | Grok 4.1 Fast (Non-reasoning) | 71.6 | #215 (Grok 4.1 Fast) |
| 10 | Claude Sonnet 4.6 | 71.1 | #90 |
| 11 | Claude 3 Opus | 70.7 | #570 |
| 12 | Claude Sonnet 4.5 | 67.6 | #142 |
| 13 | Claude Opus 4.7 | 67.2 | #47 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=moru · How It Works · Data refreshed daily, snapshot 2026-10-19.