Recent advances in high-throughput sequencing technologies have offered the possibility to generate genome-wide sequence data to delineate previously unidentified genetic structure, obtain more accurate estimates of demographic parameters, and to evaluate potential adaptive divergence. Here, we identified 27 556 single-nucleotide polymorphisms for the small yellow croaker (Larimichthys polyactis) using restriction-site-associated DNA (RAD) sequencing of 24 individuals from two populations. Significant sources of genetic variation was identified, with an average nucleotide diversity (π) of 0.00105 ± 0.000425 across individuals and long-term effective population size was thus estimated to range between 26 172 and 261 716. According to the results, no differentiation between the two populations was detected based on the SNP dataset of top quality score per contig or neutral loci. However, the two analyzed populations were highly differentiated based on SNP dataset of both top FST value per contig and the outlier SNPs. Moreover, local adaptation was highlighted by an FST-based outlier tests implemented in LOSITAN and a total of 538 potentially locally selected SNPs were identified. BLAST2GO annotation of contigs containing the outlier SNPs yielded hits for 37 (66%) of 56 significant BLASTX matches. Candidate genes for local adaptation constituted a wide array of biological functions, including cellular response to oxidative stress, actin filament binding, ion transmembrane transport and synapse assembly. The generated SNP resources in the present study provided a valuable tool for future population genetics and genomics studies of L. polyactis.