Sex is ancestral in eukaryotes. Meiotic sex differs from bacterial ways of exchanging genetic material by involving the fusion of two cells. We examine the hypothesis that fusion evolved in early eukaryotes because it was directly beneficial, rather than a passive side-effect of meiotic sex. We assume that the uptake of (proto)mitochondria into eukaryotes preceded the evolution of cell fusion, and that Muller’s ratchet operating within symbiont lineages led to the accumulation of lineage-specific sets of mutations in asexual host cells. We examine if cell fusion, and the consequent biparental inheritance of symbionts, helps to mitigate the effects of this mutational meltdown of mitochondria. In our model, host cell fitness improves when two independently-evolved mitochondrial strains co-inhabit a single cytoplasm, mirroring mitochondrial complementation found in modern eukaryotes. If fusion incurs no cost, we find that an allele coding for fusion can invade a population of non-fusers. If fusion is costly, there are two thresholds. The first describes a maximal fusing rate (probability of fusion per round of cell division) that is able to fix. An allele that codes for a rate above this threshold can reach a polymorphic equilibrium with non-fusers, as long as the rate is below the second threshold, above which the fusion allele is counterselected. Whenever it evolves, fusion increases the population-wide level of heteroplasmy, which allows mitochondrial complementation and increases population fitness. We conclude that beneficial interactions between mitochondria are a potential factor that selected for cell fusion in early eukaryotes.