Artificial intelligence (AI) disrupts the world, and dentistry is not an exception. This AI disruption creates a modern computational tool that leads to a more accurate orthodontic diagnosis. The orthodontic prediction tool, mixed dentition analysis, which is essential for the prediction of unerupted permanent canine and premolars width, now becomes possible. This prediction model enables the dental crowding assessment; therefore, it facilitates the decision-making of whether an orthodontic treatment plan requires the removal of teeth or not. Hence, the purpose of this study was to formulate and test the novel prediction of the mixed dentition analysis for Thais using AI. The sample consisted of 180 Thai individuals (90 males and 90 females) aged 11 to 30 years old. The mesiodistal widths of all maxillary and mandibular fully erupted permanent teeth, except the second and third molar teeth, were measured with pointed end digital vernier calipers. Then, the simple linear regression analysis was performed to create prediction models based on the results from the corrected Akaike information criteria (AICc). Finally, leave-one-out cross-validation (LOOCV) was used to validate these predictions. As a result, the prediction models were developed in the form of prediction equations. For example, the male’s maxillary prediction formula was Y= 0.78X + 27.40 where Y represented the predicted widths of canine and premolars and X represented the widths of mandibular central and lateral incisors. A validation as finished by LOOCV also showed that mean absolute errors (MAEs) were from 1.34 to 1.49 mm per arch. To summarize, the calculated mesiodistal width of the unerupted canine and premolars using the mandibular central and lateral incisors was proposed as another option. Within this margin of error, this novel model is considered easy and practical to predict the size of the unerupted teeth during mixed dentition for Thais.