The long-term associations between multiple metals and incident diabetes are uncertain. We aimed to examine the relationship between plasma concentrations of 23 metals and the incidence of type 2 diabetes among Chinese senior adults. We quantified fasting plasma concentrations of 23 metals by inductively coupled plasma mass spectrometry among 1039 incident diabetes cases and 1039 controls (age and sex matched) nested in a prospective study, the Dongfeng-Tongji cohort. Both cases and controls were free of diabetes at baseline (2008–2010), incident diabetes were identified using the following criteria: fasting glucose ≥ 7.0 mmoL/l; or hemoglobin A1c (HbA1c) ≥ 6.5%; or self-reported physician diagnosis of diabetes or use of anti-diabetic medication during the follow-up visits in 2013. In the conditional logistic regression models, the multivariable adjusted ORs (95% CIs) of diabetes across quartiles (Q1–Q4) of metal concentrations were as follows: titanium, 1.00, 0.92, 1.31, 1.38 (1.00–1.91, P_trend = 0.011); selenium, 1.00, 1.08, 1.45, 1.27 (0.93–1.74, P_trend = 0.05); and antimony, 1.00, 0.79, 0.77, 0.60 (0.44–0.83, P_trend = 0.002). Arsenic was significantly associated with diabetes in the crude model (ORs comparing extreme quartiles 1.30; 1.02–1.65; P_trend = 0.006), but was not significant after adjustment for socio-demographic factors. No significant associations were found for other metals. In conclusion, titanium and selenium were positively while antimony was negatively associated with incident diabetes.