A major challenge to the development of therapies for human retinal degenerative diseases is the lack of an ideal preclinical model because of the physiological differences between humans and most model animals. Despite the successful generation of a primate model through germline knockout of a disease-causing gene, the major issues restricting modeling in nonhuman primates (NTIPs) are their relatively long lifespan, lengthy gestation, and dominant pattern of singleton births. Herein, we generated three cynomolgus macaques with macular in situ knockout by subretinal delivery of an adeno-associated virus (AAV)-mediated CRISPR-Cas9 system targeting CNGB3, the gene responsible for achromatopsia. The in vivo targeting efficiency of CRISPR-Cas9 was 12%-14%, as shown by both immunohistochemistry and single-cell transcriptomic analysis. Through clinical ophthalmic examinations, we observed a reduced response of electroretinogram in the central retina, which corresponds to a somatic disruption of CNGB3. In addition, we did not detect CRISPR-Cas9 residue in the heart, liver, spleen, kidney, brain, testis, or blood a year after administration. In conclusion, we successfully generated a NHP model of cone photoreceptor dysfunction in the central retina using an in situ CNGB3-knockout strategy.