The development of advanced gene and cell therapies for the treatment of genetic diseases requires reliable animal and cellular models to test their efficacy. Moreover, the availability of the target human primary cells of these therapies is reduced in many diseases. The development of endonucleases that can cut into specific sites of the cell genome, as well as the repair of the generated break by non-homologous end-joining, results in a variety of outcomes, insertions, deletions, and inversions that can induce the disruption of any specific gene. Among the many methods that have been developed for gene editing, CRISPR-Cas9 technology has become one of the most widely used endonuclease tools due to its easy design and its low cost. It has also been reported that the use of two guides, instead of just the one required, reduces the outcomes of non-homologous end joining mainly to the precise genomic sequences between the cutting sites of the guides used. We have explored this strategy to generate useful cellular and animal models. Different distances between the two guides have been tested (from 8 to 500 bp apart), and using the optimal range of 30-60 bp we have obtained a human primary cellular model of a genetic disease, pyruvate kinase deficiency, where the availability of the target cells is limited. We have also generated an in vivo model of glycolate oxidase (GO) deficiency, which is an enzyme involved in the glyoxylate metabolism following the same strategy. We demonstrate that the use of two-guide CRISPR-Cas9-induced non-homologous end joining is a feasible and useful tool for disease modeling, and it is most relevant to those diseases in which it is difficult to get the cells that will be genetically manipulated.