Benchmarks · safety
CyberGym-E2E-AA
CyberGym-E2E-AA (based on Berkeley RDI's CyberGym-E2E), part of the AA Cyber Index: end-to-end vulnerability reproduction and patching; score. The component with the most safety refusals.
20 models · 21 results · data as of 2026-11-06 · artificialanalysis.ai ↗
Leaderboard
Kimi K3.1 Pro
More from this benchmark
4 published measuresScore vs any published measure4 measures: sub-scores, splits, counts
9 more models under All models in settings.
Breakdown
darker is better within each column · click a column to sort| Model | ||||
|---|---|---|---|---|
| Kimi K3.1 max | 839,549,087 | 1,271,256 | 2,160,314 | 820,896,283 |
| Mistral Large 4 high | 1,341,697,505 | 1,885,554 | 5,821,320 | 1,323,079,969 |
| MiMo V2.6 Pro | 1,281,494,363 | 1,833,370 | 7,786,276 | 1,259,939,237 |
| GPT-6 Luna max | 1,025,469,279 | 932,513 | 5,087,658 | 1,004,522,585 |
| DeepSeek V4.1 Flash 1017 max | 2,307,370,823 | 2,604,330 | 9,520,960 | 2,279,467,801 |
| GLM 5.3 Flash max | 1,427,617,759 | 1,924,638 | 8,166,772 | 1,407,045,636 |
| Grok 4.7 xhigh | 2,395,630,244 | 2,771,386 | 9,623,262 | 2,357,709,449 |
| Muse Spark 1.3 xhigh | 2,376,835,303 | 3,345,919 | 5,776,980 | 2,344,582,365 |
| Kimi K3 max | 767,548,156 | 1,704,231 | 5,208,358 | 752,120,036 |
| GLM 5.3 max | 576,131,552 | 835,013 | 3,638,648 | 563,012,880 |
| MiniMax M3 | 2,152,121,896 | 2,572,711 | 2,921,612 | 2,132,759,178 |
| DeepSeek V4.1 Flash max | 731,498,431 | 1,100,565 | 3,267,772 | 716,060,495 |
| Nemotron 3 Ultra max | 2,260,189,056 | 3,775,703 | 15,158,736 | 2,222,212,423 |
| Gemini 3.8 Flash high | 2,453,694,149 | 1,705,171 | 10,456,291 | 2,421,280,231 |
| Fable 5.1 max | 1,165,569,139 | 2,871,984 | 6,264,095 | 1,140,347,762 |
| Opus 5.5 max | 1,917,248,751 | 3,244,241 | 10,316,131 | 1,882,643,635 |
| GPT-6 Astra max | 248,843,171 | 1,138,351 | 2,635,316 | 232,123,904 |
| GPT-6 Sol max | 720,249,748 | 1,255,046 | 4,441,652 | 699,291,076 |
| Qwen 3.8 2.4T-A95B | 928,714,944 | 1,790,742 | 5,078,101 | 911,299,115 |
| Qwen 3.8 27B xhigh | 1,073,795,376 | 2,085,049 | 5,582,311 | 1,055,210,970 |
All results
21 rows| # | Model | Effort | CyberGym-E2E-AA |
|---|---|---|---|
| 1 | Kimi K3.1 | max | 94.7% |
| 2 | Mistral Large 4 | high | 79.4% |
| 3 | MiMo V2.6 Pro | 78.6% | |
| 4 | GPT-6 Luna | max | 77.9% |
| 5 | DeepSeek V4.1 Flash 1017 | max | 76.3% |
| 6 | GLM 5.3 Flash | max | 74.0% |
| 7 | Grok 4.7 | xhigh | 74.0% |
| 8 | Mistral Large 4 | medium | 71.8% |
| 9 | Muse Spark 1.3 | xhigh | 59.5% |
| 10 | Kimi K3 | max | 58.0% |
| 11 | GLM 5.3 | max | 29.0% |
| 12 | MiniMax M3 | 24.4% | |
| 13 | DeepSeek V4.1 Flash | max | 22.9% |
| 14 | Nemotron 3 Ultra | max | 10.7% |
| 15 | Gemini 3.8 Flash | high | 8.4% |
| 16 | Fable 5.1 | max | 0.8% |
| 17 | Opus 5.5 | max | 0.8% |
| 18 | GPT-6 Astra | max | 0.0% |
| 19 | GPT-6 Sol | max | 0.0% |
| 20 | Qwen 3.8 2.4T-A95B | 0.0% | |
| 21 | Qwen 3.8 27B | xhigh | 0.0% |