Preclinical models of cancer mutations are crucial for advancing precision oncology. Existing genetically engineered mouse models (GEMMs) represent only a minority of the various cancer alleles found in human tumours. CRISPR–Cas9 and base editors have accelerated the generation of new GEMMs, but these techniques have key limitations. CRISPR–Cas9 excels at gene disruption, but inefficiently installs templated single nucleotide variants (SNVs). SNVs comprise over 80% of somatic cancer mutations in clinical sequencing datasets, such as MSK-IMPACT. Base editors install SNVs efficiently; however, they require unique enzymes for a desired edit (for example, adenine base editors to convert A•T to G•C), generate bystander mutations, and cannot install most multinucleotide variants or insertions or deletions (indels).