Abstract. This work takes place in the context of biochemical kinetics simulation for the understanding of complex biological systems such as haemostasis. The classical approach, based on the numerical solving of differential systems, cannot satisfactorily handle local geometrical constraints, such as membrane binding events. To address this problem, we propose a particle-based system in which each molecular species is represented by a three-dimensional entity which diffuses and may undergo reactions. Such a system can be computationaly intensive, since a small time step and a very large number of entities are required to get significant results. Therefore, we propose a model that is suitable for parallel computing and that can especially take advantage of recent multicore and multiprocessor architectures. We present our particle-based system, detail the behaviour of our entities, and describe our parallel computing algorithms. Comparisons between simulations and theoretical results are exposed, as well as a performance evaluation of our algorithms.