Genetics is known to be one of the most challenging subjects in biology education because of its abstract concepts and processes. Therefore, hands-on experiments in authentic learning environments are supposed to increase comprehensibility and provide otherwise unavailable experiences to students. We applied a hands-on module in an out-of-school gene technology lab, combining experimentation and model work, in order to support the experimental work. In comparing the impact of two different approaches on cognitive achievement, cognitive load and instructional efficiency, we divided our sample (N = 254) into two groups: While both were subjected to the experimental part of the module, the "modellers" (n = 120) were required to generate a DNA model using assorted handcrafting materials, whereas the "model viewers" (n = 134) worked with a commercially available school model of DNA structure. Interestingly, the "model viewers" performed significantly better regarding a mid-term knowledge increase, while individual cognitive load scores during the activity remained similar. Accordingly, the model viewing approach produced significantly higher scores for instructional efficiency, pointing to enhanced cognitive achievement through a more intense perception of the DNA models' correct contents. While at the first glance our results seem surprising, implications for teaching when models come into play and ways to avoid such discrepancies are discussed. Consequently, recommendations for classroom impacts are presented.