Motivation: FastTree-2 is one of the most successful tools for inferring large phylogenies. With speed at the core of its design, there are still important issues in the FastTree-2 implementation that harm its performance and scalability. To deal with these limitations, we introduce VeryFastTree, a highly tuned implementation of the FastTree-2 tool that takes advantage of parallelization and vectorization strategies to boost performance. Results: VeryFastTree is able to construct a tree on a standard server using double-precision arithmetic from an ultra-large 330k alignment in only 4.5 h, which is 7.8x and 3.5x faster than the sequential and best parallel FastTree2 times, respectively.