Genome of all known retroviruses consists of two identical molecules of RNA, which are non-covalently linked. The most stable contact site between two RNA molecules is located near their 5' ends. The molecular interactions in the dimer linkage structure (DLS) in mature virions are currently unknown. Recently we suggested that the dimer linkage structure in human immunodeficiency virus 1 (HIV-1) contains both duplex and quadruplex domains and proposed a model of DLS in HIV-1Mal (Central African virus). In this paper we showed that similar models can be also built for HIV- 1Lai, a representative of the North-American and European viruses. One of the double-stranded domains in the model structures represents either an extended duplex formed by different pathways (through base pair melting and subsequent reannealing or by a recombination mechanism) or kissing loop complex. The quadruplexes contain both G- and mixed tetrads, for example, G.C.G.C or A.U.A.U. Phylogenetic analysis of 350 isolates from NCBI database showed that similar models of DLS are predictable practically for all HIV-1 isolates surveyed. A model of dimer linkage structure in Moloney murine sarcoma virus (MuSV) is also presented. The structure includes a duplex formed by the palindromic sequences and several quadruplexes.