A high nuclear grade is crucial to predicting tumor recurrence and metastasis in clear cell renal cell carcinomas (ccRCCs). We aimed to compare the mRNA profiles of tumor tissues and preoperative plasma in patients with localized T1 stage ccRCCs, and to evaluate the potential of the plasma mRNA profile for predicting high-grade ccRCCs. Data from a prospective cohort (n = 140) were collected between November 2018 and November 2019. Frozen tumor tissues and plasma were used to measure PBRM1, BAP1, SET domain-containing 2 (SETD2), KDM5C, FOXC2, CLIP4, AQP1, DDX11, BAIAP2L1, and TMEM38B mRNA levels, and correlation with the Fuhrman grade was investigated. Multivariate logistic regression analysis revealed significant association between high-grade ccRCC and SETD2 and DDX11 mRNA levels in tissues (odds ratio (β) = 0.021, 95% confidence interval (CI): 0.001–0.466, p = 0.014; β = 6.116, 95% CI: 1.729–21.631, p = 0.005, respectively) and plasma (β = 0.028, 95% CI 0.007–0.119, p < 0.001; β = 1.496, 95% CI: 1.187–1.885, p = 0.001, respectively). High-grade ccRCC prediction models revealed areas under the curve of 0.997 and 0.971 and diagnostic accuracies of 97.86% and 92.86% for the frozen tissue and plasma, respectively. SETD2 and DDX11 mRNA can serve as non-invasive plasma biomarkers for predicting high-grade ccRCCs. Studies with long follow-ups are needed to validate the prognostic value of these biomarkers in ccRCCs.