Alfalfa is a perennial, outcrossing species, and a widely planted forage legume producing highly nutritious biomass. Currently, improvement of cultivated alfalfa mainly relies on recurrent phenotypic selection. Marker assisted breeding strategies can enhance alfalfa improvement efforts, but these strategies require a large number of markers. Next generation sequencing strategies enable high-throughput discovery of SNP markers efficiently for a complex polyploid species. To increase the number of SNP for alfalfa research and molecular breeding, transcriptomes of 27 alfalfa genotypes (23 tetraploid and 4 diploid) were sequenced. De novo assembly of quality-filtered reads generated 25,183 contigs with a total length of 26.8 Mbp and an average length of 1065 bp, with an average read depth of 55.9-fold for each genotype. Overall, 21,954 (87.2%) of the 25,183 contigs represented 14,878 unique protein accessions. Gene ontology (GO) analysis suggested that a broad diversity of genes was represented in the resulting sequences. The realignment of reads to the contigs enabled the detection of 872,384 SNP and 31,760 InDels. High Resolution Melting (HRM) analysis was used to validate 91% of the 192 SNP evaluated. Structure analysis indicated that individuals from the subspecies falcata, the diploid subspecies caerulea, and the tetraploid subspecies sativa (cultivated tetraploid alfalfa) were clearly separated. The high proportion of SNP showing polymorphism among the elite tetraploid genotypes, even among the four genotypes from each breeding pool suggests the existence of high genetic variation and diversity. The EST and SNP markers generated from this study are publicly available and can contribute to future alfalfa research and breeding applications.