Objectives. Evaluate the block-adaptive number series task of reasoning, as a time-efficient proxy of general cognitive ability in the Level-2 sample of the German National Cohort (NAKO), a population-based mega cohort. Methods. The number series task consisted of two blocks of three items each, administered as part of the touchscreen-based assessment. Based on performance on the first three items, a second block of appropriate difficulty was automatically administered. Scoring of performance was based on the Rasch model. Relations of performance scores to age, sex, education, study centre, language proficiency, and scores on other cognitive tasks were examined. Results. Except for one very difficult item, the data of the remaining 14 items showed sufficient fit to the Rasch model (Infit: 0.89-1.04; Outfit: 0.80-1.08). The resulting performance scores ( N = 21,056) had a distribution that was truncated at very high levels of ability. The reliability of the performance estimates was satisfactory. Relations to age, sex, education, and the executive function factor of the other cognitive tasks in the NAKO supported the validity. Conclusions. The number series task provides a valid proxy of general cognitive ability for the Level-2 sample of the NAKO, based on a highly time-efficient assessment procedure.