LanG 2.0 Achieves 51.81% Campaign Correlation Precision Using Small Local Models
November 2, 2026
LanG 2.0 uses an agentic system with entity-anchored correlation and a tamper-evident ledger to automate security operations. It achieves 51.81% campaign correlation precision and 73.42% F1 in attack detection on unseen networks, outperforming Claude Opus 5.5 while utilizing smaller, more affordable local models.
HOW THIS AFFECTS YOU
●
builderYou can replace frontier models with smaller local models for SOC tasks by implementing structural safeguards.
●
founderThis offers a path to build highly profitable, low-latency security automation tools using cheaper compute.