Autism spectrum disorders (ASD) represent a heterogeneous group of disorders defined by deficits in social interaction/communication and restricted interests, behaviors, or activities. Models of ASD, developed based on clinical data and observations, are used in basic science, the "bench," to better understand the pathophysiology of ASD and provide therapeutic options for patients in the clinic, the "bedside." Translational medicine creates a bridge between the bench and bedside that allows for clinical and basic science discoveries to challenge one another to improve the opportunities to bring novel therapies to patients. From the clinical side, biomarker work is expanding our understanding of possible mechanisms of ASD through measures of behavior, genetics, imaging modalities, and serum markers. These biomarkers could help to subclassify patients with ASD in order to better target treatments to a more homogeneous groups of patients most likely to respond to a candidate therapy. In turn, basic science has been responding to developments in clinical evaluation by improving bench models to mechanistically and phenotypically recapitulate the ASD phenotypes observed in clinic. While genetic models are identifying novel therapeutics targets at the bench, the clinical efforts are making progress by defining better outcome measures that are most representative of meaningful patient responses. In this review, we discuss some of these challenges in translational research in ASD and strategies for the bench and bedside to bridge the gap to achieve better benefits to patients.