Very generic and free (LGPL) Belief Network Framework in C++ - supports a broad range of knowledge and dependency types for network variables and clusters of variables
Open-source suite of software tools for research and development using graphical models of probability, published by Kansas State University Laboratory for Knowledge Discovery in Databases (KDD)
Bayesian network, simulation and risk analysis software. Supports exact and approximate inference in hybrid and dynamic networks for decision support, diagnosis, statistical learning and prediction applications.