IQ-TREE is a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. The authors combine hill-climbing approaches and a stochastic perturbation method to efficiently sample local optima in the tree space. Comparative analyses with RAxML and PhyML show that IQ-TREE often achieves higher likelihoods, especially when the CPU time is limited to that required by RAxML and PhyML. However, if the CPU time is allowed to vary, IQ-TREE outperforms both programs in terms of likelihood and computational efficiency. The success of IQ-TREE is attributed to its new tree search strategy, which helps escape local optima, and the phylogenetic likelihood library, which reduces the time for likelihood computation. The authors recommend using all three programs for the best results.IQ-TREE is a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. The authors combine hill-climbing approaches and a stochastic perturbation method to efficiently sample local optima in the tree space. Comparative analyses with RAxML and PhyML show that IQ-TREE often achieves higher likelihoods, especially when the CPU time is limited to that required by RAxML and PhyML. However, if the CPU time is allowed to vary, IQ-TREE outperforms both programs in terms of likelihood and computational efficiency. The success of IQ-TREE is attributed to its new tree search strategy, which helps escape local optima, and the phylogenetic likelihood library, which reduces the time for likelihood computation. The authors recommend using all three programs for the best results.