Diabetes is a complex, multifactorial group of metabolic diseases characterized by chronic hyperglycaemia due to pancreatic beta-cell dysfunction and/or loss. It is characterized by an asymptomatic and highly variable prodromic phase, which renders diabetes mellitus difficult to be predicted with sufficient accuracy. Despite several efforts in the identification and standardization of newly trustable. Biomarkers able to predict and follow-up diabetes and to specifically subtype its different forms, few of them have proven of clinical utility. Recently, a new class of endogenous non-coding small RNAs, namely microRNAs, have been indicated as putative biomarkers, being released by cells and tissues and found in a cell-free circulating form in many biological fluids, including serum and/or plasma. MicroRNAs have been initially identified as promising biomarkers in cancer, and nowadays their application has been extended to other diseases, including diabetes. Although an increasing number of studies focused on the evaluation of circulating microRNAs in diabetes, few reproducibly identified microRNAs as biomarkers for disease prediction or follow-up. Technological problems as well as the need to obtain highly standardized operating procedures and methods are still an issue in such research field. In this review, we comprehensively resume the main and most recent findings on circulating microRNAs, and their possible use as biomarkers to predict and follow-up diabetes and its complications, as well as the methodological challenges to standardize accurate operating procedures for their analysis.