A stock assessment of the Patagonian toothfish (Dissostichus eleginoides) at Kerguelen Islands (CCAMLR Division 58.5.1) was undertaken using the modelling software CASAL, based on different statistical and biological datasets: abundance estimation from the POKER 1 (2006) and 2 (2010) surveys, standardized CPUE of commercial fisheries, tag releases and recoveries by length bin, commercial catch-at-length data... CASAL is a simple, single-sex, single-area population model. The spatial complexity in the fishery is modeled using separate fishing selectivity functions depending on the fishing zone and the type of fishing vessels (trawlers, longliners). The fishing zone for the longliners is divided into four zones (North, East, South, Skiff Bank) and into two depth zones: shallow < 1,500m; deep > 1,500m. Commercial catch are integrated in the model by CCAMLR season. A weighted combination of datasets was integrated using the effective sample sizes calculated at the level of each haul. An iteratively estimated process error was applied to the length frequency data. With an initial biomass of about 218,000 tons, the modelling tends to present results which are consistent with the stock assessments undertaken by the Australian scientific team at Heard and McDonald Islands (division 58.5.2) and over the Kerguelen Plateau. The biomass long-term projection, done for the Patagonian toothfish with a simulated 5,100 tons catch by longliners, fulfils the CCAMLR management rules. Research studies aiming at collecting more biological data and getting a better understanding of the population spatial complexity should be continued to consolidate the model.