Operational safety evaluation on OFFTOPICEVAL English 1.0 (test)
99.62AR_IDLlama-3.3 (70B)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama-3.3 (70B)Family=Llama-32025.09 | 99.62 | 69.73 | 4.21 | 53.93 | |
| Llama-3.1 (8B)Family=Llama-32025.09 | 99.52 | 25.47 | 1.62 | 23.84 | |
| GPT-OSS (120B)Family=GPT2025.09 | 99.32 | 80.42 | 35.82 | 73.33 | |
| Qwen-3 (235B)Family=Qwen2025.09 | 99.05 | 99.32 | 28.7 | 77.77 | |
| GPT-OSS (20B)Family=GPT2025.09 | 98.48 | 56.67 | 35.17 | 62.63 | |
| Phi-4 (15B)Family=Phi2025.09 | 95.14 | 83.74 | 27.75 | 70.3 | |
| Gemma-3 (12B)Family=Gemma2025.09 | 93.33 | 39.37 | 10.78 | 39.53 | |
| Qwen-3 (30B)Family=Qwen2025.09 | 84.57 | 81.89 | 23.95 | 65.1 | |
| Ministral (8B)Family=Mistral2025.09 | 74.95 | 70.09 | 5.99 | 50.47 | |
| Gemma-3 (27B)Family=Gemma2025.09 | 73.71 | 94.22 | 18.21 | 63.78 | |
| Mistral (24B)Family=Mistral2025.09 | 73.14 | 99.91 | 76.44 | 79.96 | |
| Phi-4-mini (4B)Family=Phi2025.09 | 67.24 | 35.06 | 40.95 | 48.56 |