dxFeed, a global provider of market data and financial technology solutions, announced a major upgrade to Grenadier, introducing the second generation of its AI-powered anomaly detection technology. The new implementation is powered by an upgraded asset-class agnostic model family and architected to be compatible with a broad range of market instruments. 

The enhanced release significantly expands both coverage and performance, and delivers full real-time coverage of U.S. equities order books with performance designed to continuously scan the entire symbol universe at scale.

Built on unsupervised deep learning (not LLM-based), Grenadier is engineered to uncover hidden microstructural signals associated with forthcoming volatility and market-moving events. The new generation reflects substantial work on scalability, model robustness, and production readiness for institutional environments.

Addressing Hidden Market Risks in Real Time
Electronic markets produce massive volumes of Level 2 order book data, making early detection of abnormal behavior increasingly challenging. Grenadier continuously analyzes order book states and generates normalized anomaly scores that highlight unusual structural patterns.