Abstract Summary Runs of homozygosity (ROH) are important genomic features that manifest when identical-by-descent haplotypes are inherited from parents. Their length distributions and genomic locations are informative about population history and they are useful for mapping recessive loci contributing to both Mendelian and complex disease risk. Here, we present software implementing a model-based method (Pemberton et al., 2012) for inferring ROH in genome-wide SNP datasets that incorporates population-specific parameters and a genotyping error rate as well as provides a length-based classification module to identify biologically interesting classes of ROH. Using simulations, we evaluate the performance of this method. Availability and Implementation GARLIC is written in C ++. Source code and pre-compiled binaries (Windows, OSX and Linux) are hosted on GitHub (https://github.com/szpiech/garlic) under the GNU General Public License version 3. Supplementary information Supplementary data are available at Bioinformatics online.