FastTree 2 is an improved version of the FastTree algorithm for inferring phylogenies from large sequence alignments. It enhances accuracy while maintaining scalability. FastTree 2 uses minimum-evolution subtree-pruning-regrafting (SPR) and maximum-likelihood neighbor-joining (NNI) to refine tree topologies. It also employs heuristics to reduce the search space, making it significantly faster than other methods like PhyML and RAxML. FastTree 2 estimates site-specific evolutionary rates using the CAT approximation, which is more accurate and efficient than the standard Γ4 model. It can infer phylogenies for alignments with up to 237,882 sequences in 22 hours on a desktop computer, using 5.8 GB of memory. FastTree 2 is more accurate than PhyML 3 with default settings in simulations and on real data, but less accurate than methods using SPRs. However, it is much faster, being 100–1,000 times faster for large alignments. FastTree 2 provides local support values based on the Shimodaira-Hasegawa test, which are useful for assessing tree reliability. It is suitable for bootstrapping and is available as open-source software at http://microbesonline.org/fasttree. FastTree 2 is efficient for large alignments and can handle millions of sequences, making it a valuable tool for phylogenetic analysis.FastTree 2 is an improved version of the FastTree algorithm for inferring phylogenies from large sequence alignments. It enhances accuracy while maintaining scalability. FastTree 2 uses minimum-evolution subtree-pruning-regrafting (SPR) and maximum-likelihood neighbor-joining (NNI) to refine tree topologies. It also employs heuristics to reduce the search space, making it significantly faster than other methods like PhyML and RAxML. FastTree 2 estimates site-specific evolutionary rates using the CAT approximation, which is more accurate and efficient than the standard Γ4 model. It can infer phylogenies for alignments with up to 237,882 sequences in 22 hours on a desktop computer, using 5.8 GB of memory. FastTree 2 is more accurate than PhyML 3 with default settings in simulations and on real data, but less accurate than methods using SPRs. However, it is much faster, being 100–1,000 times faster for large alignments. FastTree 2 provides local support values based on the Shimodaira-Hasegawa test, which are useful for assessing tree reliability. It is suitable for bootstrapping and is available as open-source software at http://microbesonline.org/fasttree. FastTree 2 is efficient for large alignments and can handle millions of sequences, making it a valuable tool for phylogenetic analysis.