Threat Intelligence reasoning on CyberSocEval
54.25AccuracyLLaMa-4-Maverick
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLaMa-4-Maverick2025.10 | 54.25 | 72.57 | |
| GPT-4o2025.10 | 53.57 | 73.19 | |
| LLaMa-4-Scout2025.10 | 50.34 | 69.9 | |
| CyberPal-2.0Parameters=14B2025.10 | 45.07 | 67.78 | |
| Qwen3Parameters=14B2025.10 | 43.54 | 63.12 | |
| Qwen3Parameters=8B2025.10 | 42.86 | 58.59 | |
| CyberPal-2.0Parameters=8B2025.10 | 38.61 | 65.6 | |
| CyberPal-2.0Parameters=4B2025.10 | 19.39 | 51.85 | |
| Qwen3Parameters=4B2025.10 | 5.95 | 10.12 |