Background: Rapid molecular diagnostics, with their ability to quickly identify genetic mutations associated with drug resistance in Mycobacterium tuberculosis clinical specimens, have great potential as tools to control multi- and extensively drug-resistant tuberculosis (M/XDR-TB). The Qiagen PyroMark Q96 ID system is a commercially available pyrosequencing (PSQ) platform that has been validated for rapid M/XDR-TB diagnosis. However, the details of the assay's diagnostic and technical performance have yet to be thoroughly investigated in diverse clinical environments. Methods: This study evaluates the diagnostic performance of the PSQ assay for 1128 clinical specimens from patients from three areas of high TB burden. We report on the diagnostic performance of the PSQ assay between the three sites and identify variables associated with poor PSQ technical performance. Results: In India, the sensitivity of the PSQ assay ranged from 89 to 98 % for the detection of phenotypic resistance to isoniazid, rifampicin, fluoroquinolones, and the injectables. In Moldova, assay sensitivity ranged from 7 to 94 %, and in South Africa, assay sensitivity ranged from 71 to 92 %. Specificity was high (94-100 %) across all sites. The addition of eis promoter sequencing information greatly improved the sensitivity of kanamycin resistance detection in Moldova (7 % to 79 %). Nearly all (89.4 %) sequencing reactions conducted on smear-positive, culture-positive specimens and most (70.8 %) reactions conducted on smear-negative, culture-positive specimens yielded valid PSQ reads. An investigation into the variables influencing sequencing failures indicated smear negativity, culture negativity, site (Moldova), and sequencing of the rpoB, gyrA, and rrs genes were highly associated with poor PSQ technical performance (adj. OR > 2.0). Conclusions: This study has important implications for the global implementation of PSQ as a molecular TB diagnostic, as it demonstrates how regional factors may impact PSQ diagnostic performance, while underscoring potential gene targets for optimization to improve overall PSQ assay technical performance. Trial registration: ClinicalTrials.gov ( #NCT02170441 ). Registered 12 June 2014.