Introduction. The Childhood Autism Rating Scale (CARS) is an observational instrument for assessing autism. It is the most widely used instrument for the diagnosis of autism in Mexico. Objective. To study the psychometric properties in the Mexican population using DSM-IV and DSM-5 criteria. Method. This is an observational, cross-sectional study. Participants (N = 137) were 78.8% male, with an age range between 2 and 18 years, mean 8.5. Parents completed the M-CHAT, the ABC, the ADI-R, and semi-structured clinical interviews with DSM-IV. We inferred DSM-5 criteria by selecting atypical sensory items from the ABC. Results. The internal consistency for the total CARS items was α = .88. The concurrent validity of the CARS and the DSM-IV criteria showed a Spearman coefficient of correlation of rs = .62. The convergent validity with the ADI-R showed moderate correlations, rs = .32 to .61, with the DSM-IV k = .33 and DSM-5 k = .36. The concordance with the DSM-IV and DSM-5 was 71% and 84.5%, sensitivity was 58.1% and 46-7%, and specificity 76.6% and 90.5%, respectively. The discriminant analysis showed that the CARS correctly classified 97% of children with autism, 70% with pervasive developmental disorders not otherwise specified (PDD-NOS), and 12.5% with Asperger’s syndrome. Discussion and conclusion. The Mexican version of the CARS is a valid and reliable instrument for diagnosing autism in Mexican children and adolescents.