Multi-rotor wind turbines have shown a faster wake recovery than equivalent single-rotor turbines. In this work the potential benefit of a wind farm with multi-rotor turbines is assessed through the optimization of the turbine positions using the CEGA wind farm optimization algorithm, for different power densities. The EPFL analytical wake model is used to compute the power output at each turbine. The wake growth rate in this model depends on the incoming turbulence intensity. However, added turbulence intensities caused by multi-rotors affecting downstream turbines in a wind farm have recently been shown to be different from those by single-rotors. Here, two simple expressions are derived, one for the added turbulence intensity behind a multi-rotor and another for its relationship with the wake growth rate in a downstream turbine. Results on fixed layouts show multi-rotor power output improvements between 6% (3D spacings) and 0.3% (10D spacings). This benefit is bigger under relatively low ambient turbulent levels. The optimization of a full-scale wind farm layout (Horns Rev, 80-turbines) using multi-rotors provides an advantage of 0.66% and 0.79% against the optimized and the baseline single-rotor layouts, respectively. Optimizations of power densities with average interturbine distances of 4D and 10D show 2.75% and 0.1% improvements, respectively.