This paper presents GASAT, a hybrid evolutionary algorithm for the satisfiability problem (SAT). A specific crossover operator generates new solutions, that are improved by a tabu search procedure. The performance of GASAT is assessed using a set of well-known benchmarks. Comparisons with state-of-the-art SAT algorithms show that GASAT gives very competitive results. These experiments also allow us to introduce a new SAT benchmark from a coloring problem.