We provide Stochastic Concurrent Constraint Programming (sCCP), a stochastic process algebra based on CCP, with a semantics in terms of hybrid automata. We associate with each sCCP program both a stochastic and a non-deterministic hybrid automaton. Then, we compare such automata with the standard stochastic semantics (given by a Continuous Time Markov Chain) and the one based on ordinary differential equations, obtained by a fluid-flow approximation technique. We discuss in detail two case studies: Repressilator and the Circadian Clock, with particular regard to the robustness exhibited by the different semantic models and to the effect of discreteness in dynamical evolution of such systems.