This paper develops an extended model of the classical vehicle routing problem which considers both routing and inventory costs (These costs are assumed time-dependent) and determines the optimal trade-off between these costs. Authors first formulate this problem by a mathematical programming model and an algorithm is derived using Benders' decomposition principle. This algorithm requires exact solutions and dual prices of the traveling salesman problem (TSP) using cutting plane technique. Furthermore three approximate algorithms are developed via polyhedral approximation of the TSP sub-problem. Numerical experiences show these approximate algorithms give good solutions.