The Sudoku puzzle consists in filling a square matrix with 9 rows and 9 columns, divided into 9 3 x 3 regions, so that each column, row, and region contains different digits from 1 to 9. Such a puzzle belongs to the NP- complete class of problems, existing different exact and approximate methods able to solve it. This paper reports recent results for solving Sudokus achieved by combining metaheuristics and a filtering technique coming from the constraint programming domain named AC3.