Three-dimensional (3D) technologies are increasingly used to digitize rocks, minerals, fossils, landforms, and other geological objects, but classroom evidence for their instructional value in mineral and resource geoscience remains limited. This quasi-experimental study examined whether tangible 3D-printed models were associated with higher immediate post-instruction performance than image-based instruction, and how their value varies with the representational demands of different geoscience tasks. The sample included 135 secondary students. The 3D-printing Group (n = 54) used tangible models, whereas the Traditional Group (n = 81) received image-based instruction across four topics: Cradle Stone, the Jiaodong-type Mineral System, Anchiornis huxleyi, and Detached Cave Dwellings. Baseline-adjusted analysis of covariance (ANCOVA) was used as the primary inferential test. The 3D-printing Group showed higher immediate post-test performance (adjusted means: 79.46% vs. 71.51%; difference = 7.95 percentage points; HC3 95% CI: 2.10–13.80; p = 0.008). The largest topic-specific advantage occurred for Anchiornis huxleyi, consistent with a task-dependent interpretation in which printed models are more useful when learning requires three-dimensional morphological interpretation. Questionnaire responses also favored 3D printing, especially for perceived spatial understanding. These preliminary classroom-based findings suggest that 3D-printed models are most useful when they preserve task-relevant spatial, structural, or morphological information.