As a precursor to a detailed deterministic transient analysis of the Gas-cooled Fast Reactor (GFR) and the Experimental Technology Demonstration Reactor (ETDR) options, which are currently being studied within the Generation IV GFR and EU GCFR projects, a GFR transient benchmark study was initiated. The goal of the study was to "benchmark" the transient analysis codes being used by the GFR partners: RELAPS, TRAC/AAA, CATHARE, SlM-ADS, MANTA and SPECTRA. The benchmark was coordinated by the EU GCFR project but was open to all Generation IV partners. As a starting point, the benchmark was based on a main blower failure event in ETDR with reactor scram in order to investigate the ability of the different code systems to calculate the transition in the core heat removal from the main circuit forced flow to natural circulation cooling using the plant Decay Heat Removal (DHR) system. Benchmark results were received from nine organizations: AREVA France, CEA France, NRG and TUD Netherlands, AMEC-NNC UK, INL USA, CIRTEN Italy, JRC-IE EURATOM and PSI Switzerland. For comparison purposes the benchmark was divided into several stages: the initial steady-state solution, the main blower flow run-down, the opening of the DHR loop and the transition to natural circulation and finally the "quasi" steady heat removal from the core by the DHR system. The results submitted by the participants showed that all the codes gave consistent results for all four stages of the benchmark. In the steady-state the calculations revealed some differences in the clad and fuel temperatures, the core and main loop pressure drops and in the total Helium mass inventory. Also some disagreements were observed in the Helium and water flow rates in the DHR loop during the final natural circulation stage. Good agreement was observed for the total main blower flow rate and Helium temperature rise in the core, as well as for the Helium inlet temperature into the core. In order to understand the reason for the differences in the initial "blind" calculations a second round of calculations was performed using a more precise set of boundary conditions. The paper will present the results of both sets of calculations, and highlight the differences in the different code system simulations and show how these influence the results of the benchmark calculations.