A new GPUâbased Monte Carlo dose calculation algorithm (GPUMCD), developed by the vendor Elekta for the Monaco treatment planning system (TPS), is capable of modeling dose for both a standard linear accelerator and an Elekta MRI linear accelerator. We have experimentally evaluated this algorithm for a standard Elekta Agility linear accelerator. A beam model was developed in the Monaco TPS (research version 5.09.06) using the commissioned beam data for a 6 MV Agility linac. A heterogeneous phantom representing several scenarios â tumorâinâlung, lung, and boneâinâtissue â was designed and built. Dose calculations in Monaco were done using both the current clinical Monte Carlo algorithm, XVMC, and the new GPUMCD algorithm. Dose calculations in a Pinnacle TPS were also produced using the collapsed cone convolution (CCC) algorithm with heterogeneity correction. Calculations were compared with the measured doses using an ionization chamber (A1SL) and Gafchromic EBT3 films for 2Ă2âcm2,5Ă5âcm2, and 10Ă2âcm2 field sizes. The percentage depth doses (PDDs) calculated by XVMC and GPUMCD in a homogeneous solid water phantom were within 2%/2âmm of film measurements and within 1% of ion chamber measurements. For the tumorâinâlung phantom, the calculated doses were within 2.5%/2.5âmm of film measurements for GPUMCD. For the lung phantom, doses calculated by all of the algorithms were within 3%/3âmm of film measurements, except for the 2Ă2âcm2 field size where the CCC algorithm underestimated the depth dose by âŒ5% in a larger extent of the lung region. For the bone phantom, all of the algorithms were equivalent and calculated dose to within 2%/2âmm of film measurements, except at the interfaces. Both GPUMCD and XVMC showed interface effects, which were more pronounced for GPUMCD and were comparable to film measurements, whereas the CCC algorithm showed these effects poorly.PACS number(s): 87.53.Bn, 87.55.dh, 87.55.km