Numerous applications require the precise analysis of U isotope relative enrichment in sample amounts in the subnanogram to picogram range; among those are nuclear forensics, nuclear safeguards, environmental survey, and geosciences. However, conventional thermal ionization mass spectrometry (TIMS) yields U combined ionization and transmission efficiencies (i.e., ratio of ions detected to sample atoms loaded) of less than 0.1% or 2% depending on the loading protocol, motivating the development of sources capable of enhancing ionization. The new prototype cavity source TIMS at ETH Zürich offers improvements from 4 to 15 times in combined ionization and transmission efficiency compared to conventional TIMS, yielding up to 5.6% combined efficiency. Uranium isotope ratios have been determined on reference standards in the 100 pg range bound to ion-exchange or extraction resin beads. For natural U standards, n(235U)/ n(238U) ratios are measured to relative external precisions of 0.5-1.0% (2RSD, 2 < n < 11, conventional source) or 2.0% (2RSD, n = 6, cavity source) and accuracies of 0.2-0.7% (conventional source) or 0.4-0.9% (cavity source). Meanwhile, n(234U)/ n(238U) ratios are determined to relative external precisions of 1.7-3.6% (2RSD, 2 < n < 11, conventional source) or 5.6% (2RSD, n = 6, cavity source) and accuracies of 0.1-2.5% (conventional source) or 0.5-8.3% (cavity source), which would benefit further from in-run organic interference and peak tailing corrections.