Purpose <p>Currently, relying on a single biomarker to predict the prognosis of melanoma patients is inefficient, particularly in the context of immunotherapy. We aimed to characterize tumor microenvironment (TME) subtypes and develop a robust risk classification model for melanoma prognostication.</p> Methods <p>In this study, we performed unsupervised clustering on multiple melanoma datasets and identified three distinct TME subtypes based on the expression patterns of 51 gene signatures. Furthermore, we constructed a risk model using public melanoma cohorts as the training set and validated it via multiplex immunohistochemical staining in our own tumor samples.</p> Results <p>Three unique TME subtypes (designated TME-A, -B, and -C) were identified, and these were significantly associated with overall survival, disease-specific survival and distant metastasis-free survival. Using weighted gene co-expression network analysis and the least absolute shrinkage and selection operator regression model, we further identified three pivotal genes: CXCR6, CD45, and PD-1. Based on these three genes, we developed a risk stratification model named the microenvironment subtype-related gene risk score (MSGRS), which effectively predicted the prognosis of melanoma patients across multiple independent cohorts, including populations treated with immunotherapy. Importantly, high CXCR6 expression was associated with favorable clinical outcomes, increased immune cell infiltration, and distinct spatial organization of CD8 + and CD69 + immune cells within the TME.</p> Conclusion <p>Our findings demonstrate that TME characteristics, including immune cell density and spatial distribution, and particularly the CXCR6-derived risk model, serve as robust predictors of patient survival, even in the setting of immune checkpoint inhibitor therapy.</p>