We evaluated the initial version of the Amplified Mycobacterium Tuberculosis Direct Test (Gen-Probe) (AMTDT 1) and the new version of AMTDT (AMTDT 2) for the detection of Mycobacterium tuberculosis directly from respiratory and nonrespiratory samples and compared the results with those of culture and staining methods. The assays were applied to 410 respiratory and 272 nonrespiratory samples collected from 515 patients. The combination of the culture results and clinical diagnosis was considered to be the “gold standard.” Ninety-five respiratory specimens were collected from 67 patients with a diagnosis of pulmonary tuberculosis (TB) and 68 nonrespiratory specimens were collected from 61 patients with a diagnosis of extrapulmonary TB. With respiratory specimens, the sensitivity, specificity, and positive and negative predictive values were 83, 100, 100, and 96%, respectively, for AMTDT 1 and 94.7, 100, 100, and 98.4%, respectively, for AMTDT 2. With nonrespiratory specimens, the sensitivity, specificity, and positive and negative predictive values were 83, 100, 100, and 94%, respectively, for AMTDT 1 and 86.8, 100, 100, and 98.4%, respectively, for AMTDT 2. The overall results of AMTDT 1 and AMTDT 2 were concordant for 97% (661 of 682) of the samples. Statistically significant differences in sensitivities were found between AMTDT 1 and AMTDT 2 with respiratory specimens. It was concluded that although both nucleic acid amplification methods are rapid, sensitive, and specific for the detection of M. tuberculosis complex in all types of clinical samples, AMTDT 2 appeared to be more sensitive than AMTDT 1 when applied to smear-negative specimens. In contrast AMTDT 2 is more susceptible than AMTDT 1 to inhibitory substances in the amplification reaction. The turnaround time of AMTDT 2 is shorter (3.5 h) than that for AMTDT 1 (5 h).