As the ‘great mimicker’, melanoma is characterized by a high potential for phenotypic plasticity and a wide morphologic range.1,2 In challenging cases, the correct diagnosis relies on the demonstration of a melanocytic phenotype using immunohistochemistry for S100, SOX10, Melan A/Mart-1 and, to a lesser extent, HMB-45 and/or demonstration of mutations in typical melanoma driver genes, such as BRAF, NRAS, NF1, and KIT.3,4 In addition, PRAME immunohistochemistry has become a diagnostically useful tool to help distinguish melanoma from benign melanocytic mimics.5 In a subset of melanomas, the markers of melanocytic differentiation may be partially (dedifferentiated or sarcomatoid melanoma) or completely (undifferentiated melanoma) lost. These tumors have been increasingly recognized over the past few years and may be seen in both primary and metastatic disease.1,2,6, 7, 8 In the primary setting, they pose a particular diagnostic challenge, and they closely resemble atypical fibroxanthoma (AFX) or pleomorphic dermal sarcoma (PDS). Tumors with additional heterologous differentiation and an aberrant morphologic or immunohistochemical phenotype (transdifferentiated melanoma) may easily be mistaken for rhabdomyosarcoma or spindle cell squamous cell carcinoma. Although the differential diagnosis is wide, it is typically possible to render a confident diagnosis with awareness, an index of suspicion, appropriate sampling and judicious use of immunohistochemistry and molecular diagnostics. This review article discusses our current understanding of the clinical, histologic, immunophenotypic and molecular characteristics of these rare and diagnostically challenging tumors.