HELM AIR-Bench 2024 - #6.24: Accounting: leaderboard
HELM AIR-Bench 2024 (release v1.19.0) refusal rate for the level-4 risk category "#6.24: Accounting", under level-3 "#6: Advice in Heavily Regulated Industries" 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 | O1 (2024-12-17) | 93.33 | #157 |
| 2 | O3 Mini (2025-01-31) | 86.67 | #249 |
| 3 | Claude Haiku 4.5 (20251001) | 76.67 | #264 |
| 4 | GPT-3.5 Turbo (1106) | 76.67 | #952 |
| 5 | Grok 4 | 73.33 | #189 |
| 6 | GPT-3.5 Turbo (0613) | 73.33 | #1106 |
| 7 | Granite 4.0 H Small | 66.67 | #666 |
| 8 | Granite 4.0 Micro | 66.67 | #866 |
| 9 | Grok 3 Mini Beta | 66.67 | #305 |
| 10 | GPT-5 Nano | 63.33 | #439 |
| 11 | GPT-3.5 Turbo (0125) | 63.33 | #975 |
| 12 | Gemini 2.5 Flash (Preview 04-17) | 60 | #246 |
| 13 | Gemini 2.5 Pro (Preview 03-25) | 53.33 | #126 |
| 14 | Claude 3 Opus (20240229) | 53.33 | #428 |
| 15 | GPT-4 (0613) | 53.33 | #493 |
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-2024-6-24-accounting · How It Works · Data refreshed daily, snapshot 2026-11-05.