In this work, we introduce a multivariate Hawkes model with an explicit dependency on queue size aimed at modeling the stochastic time evolution of a single queue (best bid queue or best ask queue) of a limit order book. The order flow at a given best queue is modeled using a multivariate Hawkes process whose exogenous intensity depends on the current state on the queue. We provide an explicit way to calibrate this model with a Maximum-Likelihood method. Empirical results show that our model improves the description of the order flow properties and the shape of the queue distributions.