HELM AIR-Bench - Air Bench 2024: leaderboard
The "Air Bench 2024" subtask of HELM AIR-Bench. HELM AIR-Bench: Measures model robustness, truthfulness, calibration, bias, harmfulness, jailbreak resistance, or alignment-relevant behavior.
Metric: Refusal Rate. Source: crfm.stanford.edu. Saturation forecast: Around March 2027. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Haiku 4.5 (20251001) | 93.15 | #254 |
| 2 | Claude 3.5 Sonnet (20241022) | 90.83 | #293 |
| 3 | Claude Sonnet 4.5 | 89.84 | #144 |
| 4 | Claude Sonnet 4 (20250514) | 88.27 | #218 |
| 5 | GPT-OSS-120B | 88 | #344 |
| 6 | GPT-5 Nano | 87.82 | #430 |
| 7 | GPT-5 | 87.67 | #101 |
| 8 | Qwen 3 Next 80B A3B (Thinking) | 86.7 | #323 (Qwen 3 Next 80B A3B) |
| 9 | GPT-5.1 (2025-11-13) | 86.19 | #106 |
| 10 | GPT-OSS-20B | 85.97 | #526 |
| 11 | Claude 3.5 Sonnet (20240620) | 85.9 | #300 |
| 12 | Claude Opus 4 (20250514) | 85.74 | #157 |
| 13 | GPT-5 Mini (2025-08-07) | 85.71 | #175 |
| 14 | Claude 3 Sonnet (20240229) | 84.69 | #686 |
| 15 | O3 (2025-04-16) | 84.47 | #124 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-air-bench-air-bench-2024 · How It Works · Data refreshed daily, snapshot 2026-10-24.