Simbian Research is the research arm of Simbian — publishing benchmarks, models, and tools that advance what AI can do in real threat environments.
Agentic threat hunting benchmark. An AI agent receives a threat briefing and 134K raw security events. It must find the attacker by issuing SQL queries and submit exact malicious timestamps. Deterministic scoring.
Comprehensive evaluation of frontier models across SOC tasks: alert triage, threat intelligence, log analysis, and incident response. The definitive reference for teams choosing a model for security operations.