Background Few localized food frequency questionnaires (FFQ) have been developed and used in Chinese nutrition surveys despite China’s large population and diverse dietary habits. Method We analyzed data collected in two waves (six months apart) of the Shanghai Diet and Health Study in 2012–2013, from 1623 Shanghai residents (798 men and 825 women) older than 18 years. The results of 3-day 24-h dietary recalls (HDR) plus condiment weighing were used to evaluate the validity and reliability of the SDHS FFQ. Results The median and first and third quartiles for energy intake (in kcal) derived from the FFQ1 and FFQ2 were 1566.5 (1310.1–1869.6) and 1561.9 (1280.2–1838.4), respectively, of which protein (in g) was 54.3 (42.5–65.8) and 52.9 (42.4–64.5), fat (in g) was 49.8 (37.2–64.7) and 47.9 (34.9–61.9), and carbohydrates (in g) was 227.3 (180.8–277.9) and 228.1 (182.2–275.2) in the reliability analysis. The median and first and third quartiles for energy-intake differences between the FFQ1 and the 3-day 24-HDR with condiment weighing was 59.3 (− 255.5–341.6), of which protein was − 5.2 (− 18.7–7.8) and fat was − 11.2 (− 30.8–5.3). The adjusted Spearman’s correlations were 0.33–0.77 for validity and 0.46–0.79 for reliability. The intra-class correlation coefficients exceeded 0.46 (validity) and 0.47 (reliability) for macronutrient intake. The consistency between the same and adjacent quartiles was approximately 80% for various nutrients. Conclusion The reliability and comparative validity of the SDHS FFQ is similar to FFQs that are used worldwide. Electronic supplementary material The online version of this article (10.1186/s12937-019-0454-2) contains supplementary material, which is available to authorized users.