The reproductive output of sponges possessing small elusive larvae remains little investigated, even though they are relevant community members in many cases. This makes difficult to understand their life cycles and population dynamics. In this study we monitored the dynamics of embryogenesis and larval output in a north-western Mediterranean population of the sublittoral demosponge Corticium candelabrum. The sponge was moderately abundant, mostly concentrated at overhangs and shaded walls. The population was dominated by relatively small individuals, averaging 1.26 cm3 in size. About 89% of individuals in the population produced gametes. Body size did not preclude reproduction, as gametes and embryos occurred in sponges ranging from 0.24 to 13.5 cm3 in size. On average, individuals produced 21.3 ± 12 embryos per mm3 of sponge tissue, but inter-annual variability in the reproductive effort was noticed. Larval release was asynchronous at both the individual and the population level, with sponges releasing larvae from early July to August. A yearly production was estimated of about half a million larvae per m2 of rocky bottom at those sites where this species occurs. Living larvae measured 340 ± 25 μm in length, were bulb-shaped, entirely ciliated, and swam with a clockwise rotation. They were hollow, with an internal cavity limited by a monolayered epithelium of columnar cells. All larval cells had a distal cilium arising from a principal centriole connected to a basal foot. There was an orthogonal accessory centriole giving rise to a short striated rootlet (21-nm periodicity band) that ran between the nucleus and the Golgi apparatus. This is the first reported case in Porifera of a ciliary rootlet derived from an accessory centriole, an uncommon arrangement also found in choanoflagellates. Despite sharing many features, cells of the anterior, medial, and posterior larval region were easily distinguishable from each other because of differences in their internal organization and yolk content. At the distal and proximal poles of all cells, the plasmalemma thickened to produce special junctions and seal the larval epithelium. Distal junctions were septate. The larval cavity was filled with vertically transmitted symbiotic bacteria and collagen fibrils. Collagen condensed below the larval epithelium at different degrees, but was absent in some areas, raising doubts about the occurrence of a true basement membrane. Larvae showed neither recognizable photoreceptory organelles nor evident behavioral responses to experimental light cues. Because the high larval production does not mirror adult abundance in the community, we suspect high levels of larval losses during dispersal and elevated impact of post-settlement mortality as the probable result of unselective settlement at unsuitable sites.