To diagnose gestational diabetes mellitus (GDM), plasma glucose measurements during oral glucose tolerance test (OGTT) put high demands on the methods in terms of accuracy. The aim was to evaluate and compare diagnostic performance of a point-of-care test and a glucose hexokinase laboratory method. Using risk-based screening, 175 pregnant women were included. They underwent a 75 g OGTT in their 28th (median) week of gestation. Venous blood was collected in two different tubes. Plasma glucose was measured on Cobas c701 and in duplicates on AccuChek Inform II (both methods from Roche Diagnostics). Accuracy was assessed by participating in external control programs with reference method assigned values. The methods were compared for all samples (n = 512) by regression analysis; slope of 0.90 (95% CI: 0.89-0.92), intercept of 0.12 (95% CI: 0.011-0.22) and r(s) of 0.968. The average bias between AccuChek Inform II and Cobas c701 was -8%. The proportion of women diagnosed with GDM was 25% based on AccuChek Inform II versus 55% for Cobas c701. Results from the external control program showed a bias of approximately 5% for Cobas c701 and no significant bias for AccuChek Inform II. Cobas c701 showed a large bias both towards Accu-Chek Inform II and the reference method used in the external control program, clearly exceeding the desirable bias of <2.6%. The lack of accuracy has great implications on either over- or under-diagnosis of GDM.