Malware Bench
- 01Claude Opus 4.874.2%pass
- 02GPT-5.270.9%pass
- 03Gemini 3 Pro68.1%pass
▸ leaderboards · preview
Top of each board at a glance. Open a benchmark for the full ranking, per-metric tabs, and task breakdown. Numbers shown are an illustrative preview — real results publish as each benchmark finalizes.
Long-horizon offensive operations inside realistic networks. Full kill chains that unfold over hours: recon, foothold, privilege escalation, lateral movement, exfiltration — scored on objectives reached and stealth maintained.
Defensive triage under an active adversary. Detect, contain, and remediate a running compromise while keeping production up and the blast radius small.
Deception, sandbagging, and capability elicitation. Whether a model underperforms on purpose when it senses evaluation — and whether the harness can tell.