WHBench - Harm Rate: leaderboard
Metric: Harm rate (%): share of responses failing either the unsafe-commission or the unsafe-omission safety criterion on the 47 expert-written WHBench scenarios in ten women's health topics (fertility, hormonal health, pregnancy, PCOS, contraception and others), three runs per question at temperature 0 with a 4,096-token cap, each response graded by a Claude Sonnet 4.6 judge against 4-6 clinician reference answers and a 23-criterion rubric with asymmetric safety penalties; lower is better. Source: arxiv.org. Saturation forecast: Around January 2027. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 | 12.8 |
| 2 | Claude Sonnet 4.6 | 27 |
| 3 | Mistral Large 3 | 30.5 |
| 4 | Gemini 3 Flash (Preview) | 32.6 |
| 5 | Grok 3 | 33.3 |
| 6 | Grok 4 | 37.1 |
| 7 | O3 | 38.6 |
| 8 | DeepSeek V3.2 | 44 |
| 9 | GPT-5.4 | 47.5 |
| 10 | DeepSeek R1 | 47.5 |
| 11 | Grok 3 Mini | 53.9 |
| 12 | Claude Opus 4 | 56 |
| 13 | GPT-4.1 | 61 |
| 14 | Claude Sonnet 4 | 68.1 |
| 15 | Gemini 2.5 Flash | 73.8 |
Interactive version: theaggregate.ai/benchmark?slug=whbench-harm-rate · How It Works · Data refreshed daily, snapshot 2026-10-07.