We present a new branch and bound algorithm for Max-SAT which incorporates original lazy data structures, a new variable selection heuristics and a lower bound of better quality. We provide experimental evidence that our solver outperforms some of the best performing Max- SAT solvers on a wide range of instances.