HELM AIR-Bench 2024 - #5.19: Weapon systems/Battlefield management: leaderboard
HELM AIR-Bench 2024 (release v1.19.0) refusal rate for the level-4 risk category "#5.19: Weapon systems/Battlefield management", under level-3 "#5: Autonomous Unsafe Operation of Systems" and level-2 "Operational Misuses". 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. Saturation forecast: Estimated already saturated. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 100 | #105 |
| 2 | GPT-OSS-120B | 100 | #346 |
| 3 | GPT-5 Nano | 100 | #439 |
| 4 | GPT-5 Mini (2025-08-07) | 100 | #175 |
| 5 | O3 (2025-04-16) | 100 | #125 |
| 6 | O1 (2024-12-17) | 100 | #157 |
| 7 | Claude 3 Sonnet (20240229) | 100 | #694 |
| 8 | Claude 3 Opus (20240229) | 100 | #428 |
| 9 | Claude Haiku 4.5 (20251001) | 96.67 | #264 |
| 10 | Claude 3.5 Sonnet (20241022) | 90 | #302 |
| 11 | Qwen 3 Next 80B A3B (Thinking) | 86.67 | #312 |
| 12 | Claude 3 Haiku (20240307) | 86.67 | #780 |
| 13 | Claude 3.5 Sonnet (20240620) | 83.33 | #308 |
| 14 | Claude Sonnet 4.5 | 80 | #148 |
| 15 | O4 Mini (2025-04-16) | 80 | #183 |
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-2024-5-19-weapon-systems-battlefield-management · How It Works · Data refreshed daily, snapshot 2026-11-05.