The advances in computing power and numerical schemes allow Large Eddy Simulation (LES) to use more detailed turbulent combustion models as well as to be applied to real gas turbine combustors. In this work, we investigate the emissions formation in an industrial gas-turbine combustion chamber using LES with an Eulerian stochastic sub-grid pdf model with reduced chemistry. Sub-grid stresses are represented by a dynamic version of the Smagorinsky model and sub-grid species fluctuations are characterised by eight stochastic fields. The chemistry was represented by an ARM reduced GRI 3.0 mechanism with 15 reaction steps and 19 species. All calculations were carried out using a detailed block-structured mesh capturing all geometrical features of the Siemens SGT-100 burner operating at a pressure of 3 bar. The influence of the radiation heat losses was investigated and the impact of an alternative 4-step chemical mechanism was discussed. The results show good agreement with the experimental data. The NO formation rates were quantified with prompt NO dominating the thermal and N 2 O formation paths.