Metastases are the leading cause of treatment failure and death of patients affected by malignant tumors, which makes them a major therapeutic target. During the last decade, efforts made to understand the mechanisms governing the passage of a localized tumor cell to a metastatic one has led to significant advances in this field. In vivo models using nude mice largely contributed to the understanding of this multi-step phenomenon, thus allowing the discovery of new targets and the development of therapeutic agents. These models were however hampered by the difficulties to detect micrometastases. The recent introduction of the green fluorescent protein (GFP) as a marker for tumor cells has radically changed the use of these models, in particular by allowing detection of single cell metastases in vivo. In this review, we discuss the major advantages of models using GFP-labeled cells and their limits.