This talk will describe the Graphical Models Toolkit (GMTK), a publicly available toolkit for developing graphical-model based speech and language processing systems. The talk will begin with a brief description of the representational and computational aspects of the framework. Following that will be a detailed description of GMTK's features, including a language for specifying structures and probability distributions, linear and logarithmic space training and decoding procedures, the concept of switching parents, and a generalized EM training method which allows arbitrary parameter tying both within CPTs and at the sub-Gaussian level while still achieving convergence. In doing the above, we will describe a number of graph structures that have been successfully used in speech recognition and language modeling systems. Lastly, we will describe a new algorithm for the triangulation of dynamic Graphical models (the Boundary and Partition algorithms), the use of which makes it possible to constrainedly but still optimally triangulate certain dynamic models whose optimal triangulation was unattainable using previous techniques.