Evaluation profile
AnimalHarmBench
1sub-evals
3.4%Safety weight
0%Freedom weight
1components
Weights below are portfolio-specific global index weights.
Model score (higher is better)Predicted score
About this eval
Whether model-generated text increases or decreases risk of animal harm.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction by ranking | Safety weight | Freedom weight |
|---|---|---|---|---|
| scoreahb.csv:scoreMeasures whether the model recognizes and avoids harms to non-human animals across animal-welfare scenarios. | Safety: nonhuman_ethics:1.000ahb.csv | Safety: higher | 3.4% | — |
score
Measures whether the model recognizes and avoids harms to non-human animals across animal-welfare scenarios.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | mistral-large | 0.068 | official | |
| 2 | gemini-1.5-pro | 0.066 | official | |
| 3 | gemini-1.5-flash | 0.05 | official | |
| 4 | claude-3-opus | 0.043 | official | |
| 5 | deepseek-v3 | 0.04 | official | |
| 6 | claude-3.5-haiku | 0.02 | official | |
| 7 | claude-3.5-sonnet | 0.018 | official | |
| 8 | gpt-4o | 0.011 | official | |
| 9 | gpt-4o-mini | 0.002 | official | |
| 10 | llama-3.3-70b-instruct | -0.015 | official |