HELM AIR-Bench 2024 - #14.23: Physical characteristics: leaderboard
HELM AIR-Bench 2024 (release v1.19.0) refusal rate for the level-4 risk category "#14.23: Physical characteristics", under level-3 "#14: Hate Speech (Inciting/Promoting/Expressing hatred)" and level-2 "Hate/Toxicity". The category holds only 18 prompts, so scores move in coarse steps. A GPT-4o judge grades each response from compliance (0) to refusal (1); the mean is shown as 0-100, higher meaning more refusals (safer).
Metric: Refusal Rate (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 100 | #105 |
| 2 | Claude Sonnet 4.5 | 100 | #148 |
| 3 | Qwen 2.5 72B Instruct | 100 | #469 |
| 4 | Claude 3.5 Sonnet (20241022) | 100 | #302 |
| 5 | Gemini 2.0 Flash | 100 | #351 |
| 6 | Claude 3.5 Sonnet (20240620) | 100 | #308 |
| 7 | Gemini 2.5 Pro (Preview 03-25) | 100 | #126 |
| 8 | Qwen 3 Next 80B A3B (Thinking) | 100 | #312 |
| 9 | Claude Haiku 4.5 (20251001) | 100 | #264 |
| 10 | Claude 3.7 Sonnet (20250219) | 100 | #209 |
| 11 | Claude 3 Opus (20240229) | 100 | #428 |
| 12 | Granite 4.0 H Small | 100 | #666 |
| 13 | GPT-4.5 | 100 | #278 |
| 14 | GPT-5.1 (2025-11-13) | 100 | #107 |
| 15 | Claude Sonnet 4 (20250514) | 100 | #223 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-air-bench-2024-14-23-physical-characteristics · How It Works · Data refreshed daily, snapshot 2026-11-05.