Purpose: To evaluate the surface characteristics, accuracy (trueness and precision), and dimensional stability of tooth preparation dies fabricated using conventional gypsum and direct light processing (DLP), stereolithography (SLA), and polymer jetting printing (PJP) techniques. Materials and Methods: Gypsum preparation dies were replicated according to the reference data and imported into DLP, SLA, and PJP printers, and the test data were obtained by scanning after 0, 1, 3, 7, 14, 28, and 42 days. After analyzing the surface characteristics, a best-fit algorithm between the test and the reference data was used to evaluate the accuracy and dimensional stability of the preparation dies. The data were analyzed by one-way analysis of variance and Tukey test or Kruskal-Wallis H test (α = .05). Results: Compared with the gypsum group (3.61 ± 0.59 μm), the root mean square error (RMSE) value of the SLA group (5.33 ± 0.48 μm) was rougher (P < .05), the PJP group (2.43 ± 0.37 μm) was smoother (P < .05), and the DLP group (2.92 ± 0.91 μm) had no significant difference (P > .05). For trueness, the RMSE was greater in the PJP (34.90 ± 4.91 μm) and SLA (19.01 ± 0.95 μm) groups than in the gypsum (16.47 ± 0.47 μm) group (P < .05), and no significant difference was found between the DLP (17.10 ± 1.77 µm) and gypsum groups. Regarding precision, the RMSE ranking was gypsum = DLP = SLA < PJP group. The RMSE ranges in the gypsum, DLP, PJP, and SLA groups at different times were 6.79 to 8.86 μm, 5.44 to 10.17 μm, 10.16 to 11.28 μm, and 10.94 to 32.74 μm, respectively. Conclusions: Although gypsum and 3D-printed preparation dies showed statistically significant differences in surface characteristics, accuracy, and dimensional stability, all tooth preparation dies were clinically tolerated and used to produce fixed restorations.