aiXamine - Jailbreak Robustness: leaderboard
aiXamine Jailbreak Robustness service score: enforcing alignment boundaries under PAIR, Jailbroken and cipher-encoded attacks (aiXamine, arXiv 2608.20554v1, Appendix Table 7). Normalized 0-100 score aggregated from the service's tests; higher is better. 42 models evaluated under identical black-box conditions via OpenRouter or containerized vLLM.
Metric: Service score (0-100; normalized pass rate). Source: arxiv.org. Saturation forecast: Estimated already saturated. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 27B | 99.5 | #192 |
| 2 | Qwen 3.5 122B A10B | 99.2 | #184 |
| 3 | Qwen 3.5 9B | 98.8 | #391 |
| 4 | Qwen 3.5 35B A3B | 98.7 | #268 |
| 5 | Qwen 3.5 4B | 98.6 | #514 |
| 6 | Claude Haiku 4.5 | 98.5 | #287 |
| 7 | GPT-OSS-120B | 91.9 | #346 |
| 8 | GPT-5 Mini | 90.2 | #193 |
| 9 | GPT-OSS-20B | 87.7 | #523 |
| 10 | Grok 4.1 Fast | 78.1 | #227 |
| 11 | Gemma 4 31B | 73.6 | #212 |
| 12 | Grok 4 Fast | 73.2 | #266 |
| 13 | Gemma 4 26B A4B | 72.7 | #360 |
| 14 | Claude 3.5 Haiku | 70.5 | #571 |
| 15 | Gemini 3 Flash (Preview) | 70.3 | #90 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=aixamine-jailbreak-robustness · How It Works · Data refreshed daily, snapshot 2026-11-05.