Tracheobronchial tuberculosis (TBTB) is a specific form of pulmonary tuberculosis characterized by Mycobacterium tuberculosis involvement in the tracheobronchial tree. Rapid and accurate diagnostic tools for TBTB are crucial. Raman spectroscopy (RS) is a noninvasive tool that accesses molecular vibrations and sample characteristics, enabling the creation of a molecular fingerprint of biological samples, which has enormous potential on clinical diagnosis for TBTB. This study aimed to develop and validate a diagnostic model based on RS in bronchial biopsies for identifying TBTB. The training set included patients with TBTB (n=18), airway malignant diseases (n=20), and normal mucosal tissue biopsies as the healthy controls (n=20). The spectral analysis results indicated that differential changes in tissue biomolecules, particularly certain amino acids, among the three groups. K-Nearest Neighbors (KNN), principal component analysis-linear discriminant analysis (PCA-LDA), principal component analysis-support vector machine (PCA-SVM) and decision tree methods were implemented to classify this same spectral data set. The PCA-SVM method exhibited highest classification accuracy, with a sensitivity of 88.89% and an area under the receiver operating characteristic curve (AUROC) of 0.919. Subsequently, an independent validation set comprising 121 patients with suspected TBTB was enrolled to evaluate the performance of RS model using the PCA-SVM method. The sensitivity of RS model was 87.69% (57/65) for diagnosing TBTB, higher than that of sputum smear, bronchial brush smear, the Bactec MGIT 960 Culture of bronchoalveolar lavage fluid, and comparable to GeneXpert sensitivity. In conclusion, the RS model using bronchial tissue provides a rapid and accurate method of identifying TBTB, showing potential as a powerful noninvasive tool for TBTB diagnosis under bronchoscopy.