Gas clumps formed within massive gravitationally unstable circumstellar discs are potential seeds of gas giant planets, brown dwarfs, and companion stars. Competition between three processes – migration, gas accretion, and tidal disruption – establishes what grows from a given seed. Numerical simulations and population synthesis calculations published to date, however, do not always agree on the outcome. Here, we investigate if the codes PHANTOM, GADGET, SPHINX, SEREN, GIZMO-MFM, SPHNG, and FARGO give the same answer when faced with the same migrating clump setup. Four test runs with varying assumptions about the initial clump mass and gas accretion on to it are performed. We find that the codes disagree in the clump migration rate by between 10 per cent to ∼50 per cent, depending on the test, but always arrive in the same qualitative picture. Specifically, with gas accretion turned off, planets migrate through the whole effective computational domain. In contrast, for the run with the most massive seed and gas accretion on, the planet opens a deep gap and stalls at separation of order 80 AU. We find that the artificial viscosity treatment and the sink particle prescription may account for much of the differences between the codes. We also attempt to reproduce the planet evolution tracks from our hydrodynamical simulations with prescriptions from three previous population synthesis studies. We find that the disagreement amongst the population synthesis models is far greater than that between our hydrodynamical simulations.