HELM AIR-Bench 2024 - #5.11: Heavy machinery: leaderboard
HELM AIR-Bench 2024 (release v1.19.0) refusal rate for the level-4 risk category "#5.11: Heavy machinery", under level-3 "#5: Autonomous Unsafe Operation of Systems" and level-2 "Operational Misuses" of the regulation-derived AIR 2024 taxonomy. 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: Not forecast. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Nano | 60 | #439 |
| 2 | Gemini 1.5 Flash (002) | 60 | #587 |
| 3 | Qwen 3 Next 80B A3B (Thinking) | 53.33 | #312 |
| 4 | O3 Mini (2025-01-31) | 53.33 | #249 |
| 5 | Claude 3.5 Sonnet (20241022) | 50 | #302 |
| 6 | Claude Haiku 4.5 (20251001) | 50 | #264 |
| 7 | Gemini 1.5 Pro (002) | 46.67 | #368 |
| 8 | O3 (2025-04-16) | 46.67 | #125 |
| 9 | Gemini 2.0 Pro (Preview 02-05) | 46.67 | #318 |
| 10 | GPT-3.5 Turbo (0613) | 46.67 | #1106 |
| 11 | GPT-OSS-20B | 43.33 | #523 |
| 12 | GPT-4 (0613) | 43.33 | #493 |
| 13 | Claude Sonnet 4.5 | 40 | #148 |
| 14 | Gemini 2.5 Pro (Preview 03-25) | 40 | #126 |
| 15 | GPT-5 Mini (2025-08-07) | 40 | #175 |
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-2024-5-11-heavy-machinery · How It Works · Data refreshed daily, snapshot 2026-11-05.