HELM AIR-Bench 2024 - #17.4: Sexual intercourse: leaderboard
HELM AIR-Bench 2024 (release v1.19.0) refusal rate for the level-4 risk category "#17.4: Sexual intercourse", under level-3 "#17: Adult content" and level-2 "Sexual Content" of the regulation-derived AIR 2024 taxonomy. The category holds only 15 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. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 100 | #302 |
| 2 | Gemini 2.0 Flash | 100 | #351 |
| 3 | GPT-OSS-120B | 96.67 | #346 |
| 4 | Claude 3.5 Sonnet (20240620) | 96.67 | #308 |
| 5 | Claude 3.7 Sonnet (20250219) | 96.67 | #209 |
| 6 | Qwen 2.5 72B Instruct | 93.33 | #469 |
| 7 | GPT-OSS-20B | 93.33 | #523 |
| 8 | Gemini 2.0 Pro (Preview 02-05) | 93.33 | #318 |
| 9 | GPT-4 Turbo | 90 | #417 |
| 10 | Claude Haiku 4.5 (20251001) | 90 | #264 |
| 11 | Claude Sonnet 4 (20250514) | 90 | #223 |
| 12 | Claude Opus 4 (20250514) | 90 | #164 |
| 13 | GPT-4o (2024-08-06) | 86.67 | #343 |
| 14 | Llama 3 8B Instruct | 86.67 | #1116 |
| 15 | Granite 3.3 8B Instruct | 83.33 | #815 |
Interactive version: theaggregate.ai/benchmark?slug=helm-air-bench-2024-17-4-sexual-intercourse · How It Works · Data refreshed daily, snapshot 2026-11-05.