HELM AIR-Bench 2024 - #1.4: Spear phishing: leaderboard
HELM AIR-Bench 2024 (release v1.19.0) refusal rate for the level-4 risk category "#1.4: Spear phishing", under level-3 "#1: Confidentiality" and level-2 "Security Risks" of the regulation-derived AIR 2024 taxonomy. 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 | GPT-OSS-120B | 100 | #346 |
| 4 | Claude 3.5 Sonnet (20241022) | 100 | #302 |
| 5 | O3 (2025-04-16) | 100 | #125 |
| 6 | Claude 3 Haiku (20240307) | 100 | #780 |
| 7 | Claude Opus 4 (20250514) | 100 | #164 |
| 8 | Claude 3.5 Sonnet (20240620) | 97.22 | #308 |
| 9 | GPT-OSS-20B | 94.44 | #523 |
| 10 | GPT-5 Mini (2025-08-07) | 94.44 | #175 |
| 11 | Claude Haiku 4.5 (20251001) | 94.44 | #264 |
| 12 | Claude 3 Sonnet (20240229) | 94.44 | #694 |
| 13 | GPT-5.1 (2025-11-13) | 94.44 | #107 |
| 14 | Claude Sonnet 4 (20250514) | 94.44 | #223 |
| 15 | GPT-5 Nano | 88.89 | #439 |
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-2024-1-4-spear-phishing · How It Works · Data refreshed daily, snapshot 2026-11-05.