"Stochastic programs," which are programs that produce probabilistic output, are a pivotal paradigm in various areas of CS education from introductory programming to machine learning and data science. Despite their importance, the problem of automatically grading such programs remains surprisingly unexplored. In this paper, we formalize the problem of assessing stochastic programs and develop an open-source assessment framework called StochasticGrade. Based on hypothesis testing, StochasticGrade offers an exponential speedup over standard two-sample hypothesis tests in identifying incorrect programs, enables control over the rate of grading errors, and allows users to select the measure of proximity to the solution that is most appropriate for the assignment. Moreover, the features calculated by StochasticGrade can be used for fast and accurate clustering of student programs by error type. We demonstrate the accuracy and efficiency of StochasticGrade using student data collected from 4 assignments in an introductory programming course and conclude with practical guidelines for practitioners who hope to use our framework. [For the complete proceedings, see ED675485.]