MET fusions (MET-Fs) are oncogenic drivers that remain poorly characterized. Analysis of 56 MET-F-positive tumors from an institutional cohort of 91,119 patients (79,864 DNA-seq plus 11,255 RNA-seq) uncovered two forms of MET-F pathobiology. The first group featured 5' partners with homodimerization domains fused in-frame with MET-tyrosine kinase domain (TKD), primarily originated from translocations, frequently excluded MET exon 14, mediated oncogenesis through cytoplasmic aggregation and constitutive activation, and were markedly sensitive to MET tyrosine kinase inhibitors (TKI) in pre-clinical models and patients with lung cancer. The second group lacked partner homodimerization motifs and retained MET transmembrane and extracellular domains. Their pathogenesis involved intrachromosomal rearrangements, resulting in partner selection for promoter hijacking and fusion allele amplification. Membrane-bound fusions were enriched in gliomas with RTK co-alterations. We provide a framework to comprehend the heterogeneous landscape of MET-Fs, supporting that fusion oncogenicity and MET-TKI sensitivity are determined by structural topology and pathogenomic context.