Metagenomic next-generation sequencing (mNGS) has been gradually applied to clinical practice due to its unbiased characteristics of pathogen detection. However, its diagnostic performance and clinical value in suspected pulmonary infection need to be evaluated. We systematically reviewed the clinical data of 246 patients with suspected pulmonary infection from 4 medical institutions between January 2019 and September 2021. The diagnostic performances of mNGS and conventional testing (CT) were systematically analyzed based on bronchoalveolar lavage fluid (BALF). The impacts of mNGS and CT on diagnosis modification and treatment adjustment were also assessed. The positive rates of mNGS and CT were 47.97% and 23.17%, respectively. The sensitivity of mNGS was significantly higher than that of CT (53.49% versus 23.26%, < 0.01), especially for infections of Mycobacterium tuberculosis (67.86% versus 17.86%, < 0.01), atypical pathogens (100.00% versus 7.14%, < 0.01), viruses (92.31% versus 7.69%, < 0.01), and fungi (78.57% versus 39.29%, < 0.01). The specificity of mNGS was superior to that of CT, with no statistical difference (90.32% versus 77.42%, = 0.167). The positive predictive value (PPV) and negative predictive value (NPV) of mNGS were 97.46% and 21.88%, respectively. Diagnosis modification and treatment adjustment were conducted in 32 (32/246, 13.01%) and 23 (23/246, 9.35%) cases, respectively, according to mNGS results only. mNGS significantly improved the diagnosis of suspected pulmonary infection, especially infections of M. tuberculosis, atypical pathogens, viruses, and fungi, and it demonstrated the pathogen distribution of pulmonary infections. It is expected to be a promising microbiological detection and diagnostic method in clinical practice. Pulmonary infection is a heterogeneous and complex infectious disease with high morbidity and mortality worldwide. In clinical practice, a considerable proportion of the etiology of pulmonary infection is unclear, microbiological diagnosis being challenging. Metagenomic next-generation sequencing detects all nucleic acids in a sample in an unbiased manner, revealing the microbial community environment and organisms and improving the microbiological detection and diagnosis of infectious diseases in clinical settings. This study is the first multicenter, large-scale retrospective study based entirely on BALF for pathogen detection by mNGS, and it demonstrated the superior performance of mNGS for microbiological detection and diagnosis of suspected pulmonary infection, especially in infections of Mycobacterium tuberculosis, atypical pathogens, viruses, and fungi. It also demonstrated the pathogen distribution of pulmonary infections in the real world, guiding targeted treatment and improving clinical management and prognoses.