Abstract Single-nucleus RNA sequencing (snRNA-Seq) has transformed our understanding of complex tissues, providing insights into cellular composition and heterogeneity in gene expression between cells, and their alterations in development and disease. High costs however constrain the number of samples analysed. Sample pooling and their demultiplexing following sequencing based on prior labelling with antibodies or lipid anchors conjugated to DNA barcodes (cell hashing and MULTI-seq), or using genetic differences between samples, provides a solution. However, there remains no comprehensive evaluation of these demultiplexing tools to guide selection between them. Here, we benchmark the leading software (Vireo, Souporcell, Freemuxlet, scSplit) used for sample demultiplexing using genetic variants. We further compared obtaining genetic variants from SNP array analysis of gDNA and from sample-matched bulk-RNA-Seq data, identified using three different variant calling tools (BCFtools, cellSNP, FreeBayes). Demultiplexing performance was evaluated on simulated multiplexed datasets comprising two, four, and six samples with doublet percentages between 0% and 30%, and validated against demultiplexing using sex-linked genes. Software implementation and execution were evaluated by run speed, robustness, scalability, and usability. Our results show that all tools excluding scSplit provide high recall and precision with an accuracy of 80%–85%. Vireo achieved the best accuracy. Demultiplexing tools were differentially affected by the variant calling tool with which it was paired. For all tools, accuracy decreased with the increasing percentage of doublets. Deployment of demultiplexing during analysis of pooled real-world 10x RNA-Seq data from the human heart and from different species is shown, as are advantages for doublet detection and removal.