CyberOffense - Nova
Cybersecurity
Share of the 21 objectives completed on CyberOffense's Nova range (several connected Active Directory domains, entered through a public web application), averaged over a model's runs. Objectives are sequential stages of one campaign, not independent items.
Top models (higher is better)
| Model | Score |
|---|---|
| DeepSeek V4.1 Flash 1017 | 77.8 |
| Kimi K3.1 Pro | 61.9 |
| Grok 4.7 | 57.1 |
| Kimi K3 | 42.9 |
| GLM-5.3 Flash | 23.8 |