The study analyzed radiomics based on CT images of the upper airway in 79 patients with obstructive sleep apnea (OSA) and 19 healthy controls between January 2023 and June 2024. Radio features were extracted from the CT images and processed using Python. Radiomic models were developed for OSA assessment and prediction using ten machine learning algorithms. The performance of the models was evaluated using area under the curve (AUC), calibration and decision curve analysis (DCA). The NaiveBayes algorithm achieved the best result with an AUC of 0.819 for airway, 0.812 for soft tissue and 0.854 for the full model in the test set (95% CI: 0.674–1.000). The full radiomic model performed better than the airway and soft tissue models, with satisfactory predictive calibration and clinical value in both the training and test sets. A CT-based radiomic model of the upper airway is a promising tool for OSA assessment and prediction.