OBJECTIVES: To compare the sensitivity and specificity of 1.5-T and 3.0-T diffusion-weighted MRI (DWI) to detect hyperacute ischemic stroke lesions. METHODS: We blindly reviewed the DWI of 135 acute stroke patients and 34 controls performed at 1.5 T (n = 108) or 3.0 T (n = 61). The stroke patients all had subsequently proved carotid territory ischemic stroke and were imaged within the first 6 hours after stroke onset. Four readers (2 neuroradiologists and 2 stroke neurologists) blinded to clinical data and magnetic field strength recorded the presence of ischemic lesions on DWI and apparent diffusion coefficient (ADC) maps if necessary. Sensitivity, specificity, and false-negative rates were computed. Signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and DWI contrasts were calculated at both field strengths. RESULTS: The accuracy of DWI in stroke diagnosis was superior at 1.5 T (98.8%) than at 3.0 T (90.9%, p = 0.03). The sensitivity decreased from 99.1% at 1.5 T to 92.5% at 3.0 T (p = 0.06) and the specificity from 97.8% to 84.1% (p = 0.002). ADC map readings did not improve accuracy, sensitivity, or specificity. The false-negative rate was 0.6% at 1.5 T and 6.1% at 3.0 T. Type of readers, stroke severity, and type of the coil did not affect diagnosis value. SNR and CNR were significantly higher at 3 T (p < 0.0001) but DWI contrast was lower (p = 0.04). CONCLUSIONS: Blind reading by 4 experts of a large series of images shows that 1.5-T diffusion-weighted MRI (DWI) is better than 3.0-T DWI for the imaging of hyperacute stroke during the therapeutic window of thrombolysis.