SOFA is a new open source framework primarily targeted at medical simulation research. Based on an advanced software architecture, it allows to (1) create complex and evolving simulations by combining new algorithms with algorithms already included in SOFA; (2) modify most parameters of the simulation--deformable behavior, surface representation, solver, constraints, collision algorithm, etc.--by simply editing an XML file; (3) build complex models from simpler ones using a scene-graph description; (4) efficiently simulate the dynamics of interacting objects using abstract equation solvers; and (5) reuse and easily compare a variety of available methods. In this paper we highlight the key concepts of the SOFA architecture and illustrate its potential through a series of examples.