Jailbreak Attack on 159 harmful behaviors (test)
99.37ASRCrescendoX
Evaluation Results
| Method | Links | |
|---|---|---|
| CrescendoXCompany=Mistral, Access=OW, Model=Ministral 3 8B IT, Reasoning=–2026.05 | 99.37 | |
| XTeamingCompany=Mistral, Access=OW, Model=Ministral 3 8B IT, Reasoning=–2026.05 | 96.23 | |
| CrescendoXCompany=Meta, Access=OW, Model=Llama 4 Scout, Reasoning=–2026.05 | 89.94 | |
| CrescendoXCompany=Meta, Access=OW, Model=Llama 3 70B IT, Reasoning=–2026.05 | 88.68 | |
| CrescendoXCompany=Alibaba, Access=OW, Model=Qwen 3 30B-A3B IT, Reasoning=–2026.05 | 88.68 | |
| CrescendoXCompany=NVIDIA, Access=OW, Model=Nemotron 3 Nano, Reasoning=Enabled2026.05 | 88.68 | |
| CrescendoCompany=Mistral, Access=OW, Model=Ministral 3 8B IT, Reasoning=–2026.05 | 87.42 | |
| CrescendoXCompany=Google, Access=Prop., Model=Gemini 3 Flash, Reasoning=Enabled2026.05 | 86.16 | |
| CrescendoXCompany=OpenAI, Access=Prop., Model=GPT 4o, Reasoning=–2026.05 | 84.28 | |
| CrescendoXCompany=Alibaba, Access=OW, Model=Qwen 3 4B IT, Reasoning=–2026.05 | 83.65 | |
| CrescendoXCompany=DeepSeek, Access=OW, Model=DeepSeek-V3.2, Reasoning=Enabled2026.05 | 82.39 | |
| XTeamingCompany=OpenAI, Access=Prop., Model=GPT 4o, Reasoning=–2026.05 | 81.76 | |
| CrescendoXCompany=xAI, Access=Prop., Model=Grok 4.1 Fast, Reasoning=Enabled2026.05 | 81.13 | |
| CrescendoXCompany=Google, Access=OW, Model=Gemma 4 E4B IT, Reasoning=–2026.05 | 78.62 | |
| CrescendoCompany=OpenAI, Access=Prop., Model=GPT 4o, Reasoning=–2026.05 | 78.62 | |
| CrescendoXCompany=Meta, Access=OW, Model=Llama 4 Maverick, Reasoning=–2026.05 | 77.36 | |
| CrescendoCompany=Alibaba, Access=OW, Model=Qwen 3 30B-A3B IT, Reasoning=–2026.05 | 74.21 | |
| XTeamingCompany=Meta, Access=OW, Model=Llama 3 70B IT, Reasoning=–2026.05 | 74.17 | |
| CrescendoXCompany=Google, Access=OW, Model=Gemma 4 26B-A4B IT, Reasoning=–2026.05 | 73.42 | |
| XTeamingCompany=DeepSeek, Access=OW, Model=DeepSeek-V3.2, Reasoning=Enabled2026.05 | 72.96 | |
| CrescendoCompany=Google, Access=Prop., Model=Gemini 3 Flash, Reasoning=Enabled2026.05 | 72.96 | |
| CrescendoXCompany=Meta, Access=OW, Model=Llama 3 8B IT, Reasoning=–2026.05 | 71.07 | |
| CrescendoCompany=Meta, Access=OW, Model=Llama 3 70B IT, Reasoning=–2026.05 | 68.55 | |
| XTeamingCompany=Alibaba, Access=OW, Model=Qwen 3 4B IT, Reasoning=–2026.05 | 67.74 | |
| CrescendoXCompany=Alibaba, Access=Prop., Model=Qwen 3.5 Flash, Reasoning=Enabled2026.05 | 65.41 | |
| CrescendoCompany=NVIDIA, Access=OW, Model=Nemotron 3 Nano, Reasoning=Enabled2026.05 | 65.41 | |
| CrescendoCompany=Meta, Access=OW, Model=Llama 4 Scout, Reasoning=–2026.05 | 64.78 | |
| CrescendoCompany=DeepSeek, Access=OW, Model=DeepSeek-V3.2, Reasoning=Enabled2026.05 | 64.78 | |
| CrescendoCompany=Meta, Access=OW, Model=Llama 4 Maverick, Reasoning=–2026.05 | 63.52 | |
| CrescendoXCompany=Google, Access=Prop., Model=Gemini 3 Pro, Reasoning=Enabled2026.05 | 63.52 | |
| CrescendoCompany=Alibaba, Access=OW, Model=Qwen 3 4B IT, Reasoning=–2026.05 | 62.89 | |
| CrescendoCompany=Google, Access=Prop., Model=Gemini 3 Pro, Reasoning=Enabled2026.05 | 62.89 | |
| XTeamingCompany=Alibaba, Access=OW, Model=Qwen 3 30B-A3B IT, Reasoning=–2026.05 | 62.42 | |
| XTeamingCompany=xAI, Access=Prop., Model=Grok 4.1 Fast, Reasoning=Enabled2026.05 | 62.26 | |
| XTeamingCompany=Meta, Access=OW, Model=Llama 4 Scout, Reasoning=–2026.05 | 59.75 | |
| CrescendoCompany=Google, Access=OW, Model=Gemma 4 26B-A4B IT, Reasoning=–2026.05 | 58.49 | |
| CrescendoCompany=Meta, Access=OW, Model=Llama 3 8B IT, Reasoning=–2026.05 | 57.86 | |
| XTeamingCompany=NVIDIA, Access=OW, Model=Nemotron 3 Nano, Reasoning=Enabled2026.05 | 57.86 | |
| XTeamingCompany=Meta, Access=OW, Model=Llama 4 Maverick, Reasoning=–2026.05 | 55.97 | |
| CrescendoXCompany=NVIDIA, Access=OW, Model=Nemotron 3 Super, Reasoning=Enabled2026.05 | 55.97 | |
| XTeamingCompany=Google, Access=Prop., Model=Gemini 3 Flash, Reasoning=Enabled2026.05 | 53.46 | |
| CrescendoXCompany=Alibaba, Access=Prop., Model=Qwen 3.5 Plus, Reasoning=Enabled2026.05 | 52.2 | |
| CrescendoCompany=xAI, Access=Prop., Model=Grok 4.1 Fast, Reasoning=Enabled2026.05 | 49.69 | |
| CrescendoCompany=Google, Access=OW, Model=Gemma 4 E4B IT, Reasoning=–2026.05 | 48.43 | |
| XTeamingCompany=Google, Access=OW, Model=Gemma 4 26B-A4B IT, Reasoning=–2026.05 | 47.8 | |
| XTeamingCompany=Meta, Access=OW, Model=Llama 3 8B IT, Reasoning=–2026.05 | 47.77 | |
| XTeamingCompany=Google, Access=OW, Model=Gemma 4 E4B IT, Reasoning=–2026.05 | 46.45 | |
| CrescendoCompany=NVIDIA, Access=OW, Model=Nemotron 3 Super, Reasoning=Enabled2026.05 | 42.14 | |
| XTeamingCompany=Google, Access=Prop., Model=Gemini 3 Pro, Reasoning=Enabled2026.05 | 40.25 | |
| XTeamingCompany=Alibaba, Access=Prop., Model=Qwen 3.5 Plus, Reasoning=Enabled2026.05 | 39.62 | |
| CrescendoXCompany=OpenAI, Access=Prop., Model=GPT 5, Reasoning=Enabled2026.05 | 38.71 | |
| XTeamingCompany=NVIDIA, Access=OW, Model=Nemotron 3 Super, Reasoning=Enabled2026.05 | 37.74 | |
| CrescendoCompany=Alibaba, Access=Prop., Model=Qwen 3.5 Plus, Reasoning=Enabled2026.05 | 36.48 | |
| XTeamingCompany=OpenAI, Access=Prop., Model=GPT 5, Reasoning=Enabled2026.05 | 32.7 | |
| CrescendoCompany=Anthropic, Access=Prop., Model=Claude Sonnet 4.5, Reasoning=Enabled2026.05 | 30.82 | |
| CrescendoCompany=Alibaba, Access=Prop., Model=Qwen 3.5 Flash, Reasoning=Enabled2026.05 | 28.3 | |
| CrescendoCompany=OpenAI, Access=Prop., Model=GPT 5, Reasoning=Enabled2026.05 | 25.16 | |
| CrescendoXCompany=Anthropic, Access=Prop., Model=Claude Sonnet 4.5, Reasoning=Enabled2026.05 | 24.53 | |
| XTeamingCompany=Alibaba, Access=Prop., Model=Qwen 3.5 Flash, Reasoning=Enabled2026.05 | 23.27 | |
| CrescendoCompany=Anthropic, Access=Prop., Model=Claude Haiku 4.5, Reasoning=–2026.05 | 21.38 | |
| CrescendoXCompany=Anthropic, Access=Prop., Model=Claude Haiku 4.5, Reasoning=–2026.05 | 18.24 | |
| XTeamingCompany=Anthropic, Access=Prop., Model=Claude Sonnet 4.5, Reasoning=Enabled2026.05 | 3.14 | |
| XTeamingCompany=Anthropic, Access=Prop., Model=Claude Haiku 4.5, Reasoning=–2026.05 | 0.63 |