HELM AIR-Bench 2024 - #39-40.51: Classification of individuals - Social behaviors: leaderboard
HELM AIR-Bench 2024 (v1.19.0) refusal rate for level-4 risk category "#39-40.51: Classification of individuals - Social behaviors" (level-3 "#39-40: Discrimination/Protected Characteristics Combinations", level-2 "Discrimination/Bias"). Only 30 prompts, so scores move in coarse steps. A GPT-4o judge scores each response from compliance (0) to refusal (1); shown as 0-100, higher means 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 | Gemini 1.5 Pro (002) | 100 | #368 |
| 2 | Qwen 3 Next 80B A3B (Thinking) | 100 | #312 |
| 3 | O3 Mini (2025-01-31) | 100 | #249 |
| 4 | Claude Opus 4 (20250514) | 100 | #164 |
| 5 | GPT-OSS-120B | 98.33 | #346 |
| 6 | GPT-5 Nano | 98.33 | #439 |
| 7 | Gemini 3 Pro (Preview) | 98.33 | #71 |
| 8 | Gemini 1.5 Flash (002) | 98.33 | #587 |
| 9 | GPT-OSS-20B | 98.33 | #523 |
| 10 | Claude Haiku 4.5 (20251001) | 98.33 | #264 |
| 11 | Grok 3 | 98.33 | #313 |
| 12 | O3 (2025-04-16) | 98.33 | #125 |
| 13 | GPT-5.1 (2025-11-13) | 98.33 | #107 |
| 14 | Gemini 2.5 Flash (Preview 04-17) | 98.33 | #246 |
| 15 | Claude 3.5 Sonnet (20241022) | 96.67 | #302 |
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-2024-39-40-51-classification-of-individuals-social-behaviors · How It Works · Data refreshed daily, snapshot 2026-11-05.