aiXamine - Adversarial Robustness: leaderboard
aiXamine Adversarial Robustness service score: resistance to adversarial prompt perturbations on AdvGLUE and AdvGLUE++ tasks (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: Around February 2027. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 77.2 | #90 |
| 2 | Gemma 4 31B | 75.2 | #212 |
| 3 | GPT-5 Mini | 71.6 | #193 |
| 4 | Grok 4 Fast | 70.3 | #266 |
| 5 | Gemma 4 26B A4B | 70.1 | #360 |
| 6 | Gemma 2 27B | 69.9 | #803 |
| 7 | Qwen 3.5 27B | 69.8 | #192 |
| 8 | Qwen 3.5 122B A10B | 69.2 | #184 |
| 9 | Gemma 3 12B | 68.6 | #674 |
| 10 | Grok 4.1 Fast | 68.3 | #227 |
| 11 | Qwen 3.5 35B A3B | 68.1 | #268 |
| 12 | Llama 3.3 70B | 67.6 | #578 |
| 13 | Gemma 3 27B | 67.3 | #606 |
| 14 | DeepSeek R1 0528 | 67.2 | #232 |
| 15 | GPT-OSS-120B | 66.3 | #346 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=aixamine-adversarial-robustness · How It Works · Data refreshed daily, snapshot 2026-11-05.