HELM AIR-Bench 2024 - #39-40.29: Characterization of identity - Geographic region: leaderboard
HELM AIR-Bench 2024 (v1.19.0) refusal rate for level-4 risk category "#39-40.29: Characterization of identity - Geographic region" (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 | GPT-5 | 100 | #105 |
| 2 | Claude Sonnet 4.5 | 100 | #148 |
| 3 | GPT-5 Nano | 100 | #439 |
| 4 | GPT-5 Mini (2025-08-07) | 100 | #175 |
| 5 | GPT-5.1 (2025-11-13) | 100 | #107 |
| 6 | Claude Haiku 4.5 (20251001) | 98.33 | #264 |
| 7 | O1 (2024-12-17) | 95 | #157 |
| 8 | Claude Sonnet 4 (20250514) | 95 | #223 |
| 9 | O3 (2025-04-16) | 93.33 | #125 |
| 10 | Claude Opus 4 (20250514) | 93.33 | #164 |
| 11 | GPT-OSS-120B | 91.67 | #346 |
| 12 | Claude 3.5 Sonnet (20240620) | 90 | #308 |
| 13 | Gemini 2.5 Pro (Preview 03-25) | 90 | #126 |
| 14 | GPT-4 Turbo | 90 | #417 |
| 15 | Qwen 3 Next 80B A3B (Thinking) | 90 | #312 |
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-2024-39-40-29-characterization-of-identity-geographic-region · How It Works · Data refreshed daily, snapshot 2026-11-05.