The dataset contains carbon (C) and nitrogen (N) stable isotope compositions analysed in the muscle tissue of 15 meso- to bathypelagic species sampled in the twilight zone (deep pelagic area) of the Bay of Biscay, North-East Atlantic. The species included 4 crustacean species (Pasiphaea sivado, Sergia robusta, Systellaspis debilis, Ephyrina figueirai) and 11 fish species (Xenodermichthys copei, Searsia koefoedi, Myctophum punctatum, Notoscopelus kroeyeri, Lampanyctus crocodilus, Argyropelecus olfersii, Arctozenus risso, Stomias boa, Serrivomer beanii, Chauliodus sloani, Aphanopus carbo). Specimens were collected during a single fishery in a canyon of the slope of the Bay of Biscay in October 2017, during the EVHOE fishery survey (“Evaluation Halieutique de l'Ouest de l'Europe”; https://doi.org/10.17600/17002300) conducted each autumn by the “Institut Français de Recherche pour l'Exploitation de la Mer” (Ifremer) on R/V Thalassa. A total of 266 individuals belonging to the 15 species were collected at night using a 25 m vertical opening pelagic trawl in the deep scattering layer (ca. 800 m depth in the water column; 1330 m bottom depth). All organisms were collected during one haul of 60 min, at a speed of approximately 4 knots (geographical coordinates at the beginning of the turn/end of the fishing: 45.103°N, -3.543° W).For small fish and crustaceans, organisms belonging to the same species were pooled by individuals of similar sizes. The size of each individual (total length for fish, cephalothorax length for crustaceans, in mm) as well as the total fresh weight of individuals or pools (to the nearest 0.5 g wet mass) were determined on board, and the individuals were rinsed with ultrapure water before storage. Mean individual sizes and fresh wet weights are here reported for each sample constituted by a pool of individuals. Samples (individuals or pools of individuals, N=39 in total) were finally stored at -20°C until further treatment in the laboratory.In clean and contamination-free conditions of the laboratory, whole organisms were briefly thawed and a small piece of white muscle (typically <3% of individual total weight) was collected from each individual. The muscle tissue is indeed generally recommended in the literature for food web studies inferred from stable isotope analyses (Pinnegar and Polunin, 1999). After collection, muscle subsamples were frozen again at -20°C, freeze-dried and homogenised manually into a fine powder. An aliquot of this powder (0.40 ± 0.05 mg dry mass) was weighed in tin cups. Analyses were finally performed with an isotope ratio mass spectrometer (Delta V Advantage with a Conflo IV interface, Thermo Scientific) coupled to an elemental analyser (Flash EA 2000, Thermo Scientific). The results are presented in the usual δ notation relative to the deviation from international standards (Vienna Pee Dee Belemnite for δ13C values, and atmospheric nitrogen for δ15N values), in parts per thousand (‰). Based on replicate measurements of USGS-61 and USGS-62 used as laboratory internal standards, experimental analytical precision was <0.10‰ and <0.15‰ for δ13C and δ15N, respectively. With the elemental analyser, bulk C:N ratios in muscle could be also determined as a proxy of the lipid content or body condition of organisms (Hoffman et al., 2015; Post et al., 2007). Samples were thus untreated (not lipid-extracted) before analyses in order to have access to bulk (untreated) C:N ratios. However, lipids are highly depleted in 13C relative to other tissue components (DeNiro and Epstein, 1977) and significant variations in lipids (especially between species) can affect δ13C signatures even if trophic sources are similar. Before using data as trophic markers, we thus recommend to mathematically correct δ13C values for the potential effect of lipids according to the formula proposed by Post et al. (2007) using bulk C:N ratios (δ13C (corrected) = δ13C (bulk) – 3.32 + 0.99 x C:N ratio). Alternatively, δ15N values do not need to be corrected.