OBJECTIVE To study the diagnostic value and pitfalls of ultrasound-guided core needle biopsy (CNB) of soft tissue tumors. METHODS One hundred and six cases of CNB specimens encountered during the period from 2007 to 2012 were enrolled into the study. The pathologic diagnosis using CNB was compared with that using surgical specimens. Diagnostic accuracy was analyzed using Chi-square test, with respect to the histologic pattern (such as spindle cell and myxoid), biologic behavior (benign versus malignant) and immunohistochemical results. The 59 cases of sarcoma were subdivided into three grades according to FNCLCC grading system. RESULTS Histologic diagnosis could be made in 84.0% (89/106) cases. Thirteen cases were non-diagnostic on CNB. There were 4 cases on CNB showing diagnostic discrepancy with surgical specimens. Four cases of "benign lesions" on CNB found to be myxoid liposarcoma and lipoma-like liposarcoma upon resection. In general, myxoid pattern (9/17) seen on CNB showed less diagnostic correlation with surgical specimens, as compared to spindle cell and other histologic patterns (P < 0.01). The rate of diagnostic correlation was 79.7% (49/59) for the 59 cases of sarcoma studied, with grade 2 and grade 3 sarcoma showing better correlation (in contrast to 7/17 for grade 1 sarcoma) (P < 0.01). Comparative analysis showed no significant difference between benign/borderline tumors and sarcomas. The application of immunohistochemical study did not result in significant improvement in diagnostic accuracy on CNB. CONCLUSIONS Ultrasound-guided CNB is a reliable tool in pathologic diagnosis of soft tissue tumors and shows a high accuracy rate especially for high-grade sarcoma. Tumors with myxoid pattern, lipomatous tumors and grade 1 sarcomas are associated with lower diagnostic accuracy on CNB. Correlation with clinicoradiologic findings would also be helpful in diagnostic evaluation and surgical planning.