Prompt Injection
Measures the model's performance against known injection attacks. (Higher score is better.)
| Rank | Model | Provider | ||||
|---|---|---|---|---|---|---|
| #1 | Claude 4.5 Haiku | Anthropic | 98.07% | 97.86% | 97.59% | 98.75% |
| #2 | Claude 4.1 Opus | Anthropic | 97.75% | 97.86% | 97.89% | 97.49% |
| #3 | Claude 4.5 Sonnet | Anthropic | 97.42% | 97.33% | 98.19% | 96.74% |
| #4 | Claude 4.5 Opus | Anthropic | 97.05% | 97.33% | 97.59% | 96.24% |
| #5 | Claude 4.6 Opus | Anthropic | 95.35% | 97.31% | 96.07% | 92.68% |
| #6 | Claude 3.5 Haiku 20241022 | Anthropic | 93.40% | 93.58% | 91.87% | 94.74% |
| #7 | Claude 5 Sonnet | Anthropic | 93.30% | 96.79% | 93.37% | 89.72% |
| #8 | Claude 4.6 Sonnet | Anthropic | 92.43% | 91.89% | 92.73% | 92.68% |
| #9 | GPT 5.2 | OpenAI | 89.02% | 92.41% | 87.25% | 87.39% |
| #10 | GPT 5.5 | OpenAI | 88.66% | 93.55% | 89.02% | 83.42% |
| #11 | Qwen 3.7 Max | Alibaba Qwen | 88.16% | 88.14% | 90.63% | 85.71% |
| #12 | Kimi K2.6 | Moonshot AI | 87.60% | 90.37% | 86.45% | 85.96% |
| #13 | GPT 5 mini | OpenAI | 86.55% | 90.22% | 87.16% | 82.28% |
| #14 | Claude 3.7 Sonnet | Anthropic | 86.35% | 87.70% | 86.14% | 85.21% |
| #15 | GPT 5 nano | OpenAI | 86.17% | 90.37% | 84.94% | 83.21% |
| #16 | Qwen 3.7 Plus | Alibaba Qwen | 86.06% | 87.57% | 84.89% | 85.71% |
| #17 | GPT OSS 120B | OpenAI | 86.00% | 92.51% | 84.04% | 81.45% |
| #18 | Kimi K2.5 | Moonshot AI | 85.21% | 89.84% | 82.83% | 82.96% |
| #19 | GPT 5.1 | OpenAI | 85.19% | 89.84% | 84.04% | 81.70% |
| #20 | GPT 5 | OpenAI | 81.30% | 85.95% | 79.82% | 78.14% |
| #21 | Llama 3.1 405B Instruct OR | Meta | 79.06% | 78.61% | 84.64% | 73.93% |
| #22 | Grok 4 | xAI | 77.16% | 79.14% | 74.40% | 77.94% |
| #23 | GPT 4.1 nano | OpenAI | 75.85% | 82.89% | 70.48% | 74.19% |
| #24 | GLM 5.2 | Z.ai | 75.67% | 83.05% | 72.29% | 71.68% |
| #25 | Grok 4.3 | xAI | 75.20% | 78.07% | 77.11% | 70.43% |
| #26 | Gemma 4 | Google | 71.23% | 76.47% | 71.30% | 65.91% |
| #27 | GPT 4o | OpenAI | 71.17% | 75.40% | 68.67% | 69.42% |
| #28 | GPT 4.1 | OpenAI | 69.98% | 79.68% | 63.86% | 66.42% |
| #29 | Gemini 3.1 Pro Preview | Google | 69.11% | 75.94% | 67.47% | 63.91% |
| #30 | Qwen Plus | Alibaba Qwen | 68.88% | 79.21% | 66.27% | 61.15% |
| #31 | Llama 4 Maverick | Meta | 68.33% | 61.50% | 79.82% | 63.66% |
| #32 | Gemini 3.0 Pro Preview | Google | 67.99% | 77.01% | 66.57% | 60.40% |
| #33 | GPT 4o mini | OpenAI | 67.28% | 73.26% | 63.86% | 64.74% |
| #34 | Qwen 2.5 Max | Alibaba Qwen | 66.73% | 77.53% | 59.52% | 63.16% |
| #35 | DeepSeek V4 Flash | Deepseek | 66.05% | 56.15% | 72.59% | 69.42% |
| #36 | DeepSeek V4 Pro | Deepseek | 64.69% | 50.27% | 71.39% | 72.43% |
| #37 | Qwen 3 Max | Alibaba Qwen | 64.44% | 72.47% | 61.45% | 59.40% |
| #38 | GPT 4.1 mini | OpenAI | 61.33% | 69.35% | 60.24% | 54.39% |
| #39 | Grok 4 Fast No Reasoning | xAI | 60.31% | 63.10% | 59.94% | 57.89% |
| #40 | Gemini 3.1 Flash Lite | Google | 60.31% | 70.05% | 60.24% | 50.63% |
| #41 | Gemini 3.5 Flash | Google | 56.66% | 62.03% | 57.83% | 50.13% |
| #42 | Llama 4 Scout | Meta | 55.76% | 52.94% | 64.46% | 49.87% |
| #43 | Deepseek R1 0528 | Deepseek | 54.41% | 47.59% | 60.24% | 55.39% |
| #44 | Qwen 3 30B VL Instruct | Alibaba Qwen | 53.93% | 63.64% | 50.90% | 47.24% |
| #45 | Gemini 2.5 Flash Lite | Google | 53.18% | 57.22% | 52.71% | 49.62% |
| #46 | Deepseek V3.1 | Deepseek | 52.52% | 48.66% | 54.52% | 54.39% |
| #47 | Gemini 2.5 Flash | Google | 50.14% | 54.55% | 51.51% | 44.36% |
| #48 | Llama 3.1 8B Instruct | Meta | 49.57% | 51.34% | 53.78% | 43.61% |
| #49 | Gemini 2.5 Pro | Google | 48.59% | 54.01% | 46.39% | 45.36% |
| #50 | Llama 3.3 70B Instruct OR | Meta | 48.10% | 40.64% | 58.43% | 45.23% |
| #51 | Gemini 2.0 Flash | Google | 46.15% | 47.59% | 45.76% | 45.09% |
| #52 | Gemini 2.0 Flash Lite | Google | 46.13% | 52.41% | 39.88% | 46.10% |
| #53 | Gemma 3 12B IT OR | Google | 45.86% | 47.06% | 42.77% | 47.74% |
| #54 | Deepseek V3 0324 | Deepseek | 44.98% | 47.06% | 45.78% | 42.11% |
| #55 | Qwen 3 8B | Alibaba Qwen | 44.33% | 45.45% | 43.94% | 43.61% |
| #56 | Gemma 3 27B IT OR | Google | 41.97% | 36.36% | 48.19% | 41.35% |
| #57 | Magistral Medium Latest | Mistral | 41.33% | 37.43% | 52.71% | 33.83% |
| #58 | Mistral Medium 3.5 | Mistral | 40.60% | 45.45% | 42.17% | 34.17% |
| #59 | Command A | Cohere | 39.97% | 36.90% | 40.66% | 42.36% |
| #60 | Deepseek V3 | Deepseek | 38.19% | 39.57% | 39.16% | 35.84% |
| #61 | Grok 2 | xAI | 35.57% | 35.14% | 36.97% | 34.60% |
| #62 | Mistral Small 3.2 | Mistral | 34.20% | 33.69% | 34.34% | 34.59% |
| #63 | Grok 3 | xAI | 32.99% | 33.16% | 35.24% | 30.58% |
| #64 | Mistral Medium Latest | Mistral | 31.00% | 28.34% | 36.45% | 28.21% |
| #65 | Mistral Large 2 | Mistral | 30.58% | 27.27% | 36.14% | 28.32% |
| #66 | Grok 3 mini | xAI | 28.72% | 24.60% | 33.73% | 27.82% |
| #67 | Mistral Large 3 | Mistral | 24.56% | 23.53% | 25.60% | 24.56% |
| #68 | Magistral Small Latest | Mistral | 22.09% | 17.11% | 27.11% | 22.06% |
| Mistral Small 3.1* | Mistral | N/A | N/A | N/A | N/A | |
| Claude 3.5 Sonnet* | Anthropic | N/A | N/A | N/A | N/A | |
| Gemini 1.5 Pro* | Google | N/A | N/A | N/A | N/A |
* Models marked with an asterisk have partial scores.