HELM AIR-Bench 2024 - #39-40.22: Characterization of identity - Ethnicity: leaderboard
Metric: Refusal Rate (%). Source: crfm.stanford.edu. 87 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 96.67 |
| 2 | Claude Sonnet 4.5 | 96.67 |
| 3 | Claude 3.5 Sonnet (20241022) | 91.67 |
| 4 | Gemini 2.5 Pro (Preview 03-25) | 91.67 |
| 5 | Claude Sonnet 4 (20250514) | 91.67 |
| 6 | Claude Haiku 4.5 (20251001) | 90 |
| 7 | GPT-5 Mini (2025-08-07) | 88.33 |
| 8 | Gemini 2.5 Flash (Preview 04-17) | 88.33 |
| 9 | Gemini 1.5 Pro (002) | 86.67 |
| 10 | GPT-5 Nano | 86.67 |
| 11 | Qwen 3 Next 80B A3B (Thinking) | 86.67 |
| 12 | O3 Mini (2025-01-31) | 85 |
| 13 | Claude Opus 4 (20250514) | 85 |
| 14 | Claude 3.5 Sonnet (20240620) | 81.67 |
| 15 | Gemini 1.5 Flash (002) | 80 |
Interactive version: theaggregate.ai/benchmark?slug=helm-air-bench-2024-39-40-22-characterization-of-identity-ethnicity · How It Works · Data refreshed daily, snapshot 2026-09-19.