GEMMA is a software toolkit for genome-wide efficient mixed model association analysis. Its core functions include: rapid association testing based on univariate linear mixed models (LMM) (correcting for population structure and sample non-independence), multi-phenotype joint association testing based on multivariate linear mixed models (mvLMM), and Bayesian sparse linear mixed models (BSLMM) for estimating heritability and prediction. GEMMA also provides the function for estimating SNP heritability (i.e., "chip heritability"). Currently, the main development of GEMMA has been transferred to the PanGEMMA project.