This paper presents a benchmarking study of non-adiabatic quantum dynamics using the molecular Tully models. The authors validate the on-the-fly direct dynamics variational multi-configuration Gaussian (DD-vMCG) method using three molecular systems: ethene (IC1), DMABN (IC2), and fulvene (IC3). These systems were proposed by Ibele and Curchod as molecular versions of the Tully models used to test one-dimensional non-adiabatic behaviour. The study compares the performance of DD-vMCG with other methods, including Tully Surface Hopping (TSH) and Ab Initio Multiple Spawning (AIMS). The authors also test parametrised linear vibronic potential energy surfaces for each system and compare them to on-the-fly results. The three systems exhibit different deactivation pathways after excitation to their ππ* bright states. When comparing DD-vMCG to AIMS and TSH, the authors find crucial differences in some cases, which are explained by the classical nature and initial conditions of the TSH simulations. The study also explores the characterisation of conical intersections, which are classified as either peaked or sloped. The authors find that the conical intersection of DMABN is peaked, while the population in fulvene passes through a sloped conical intersection. The paper also discusses the methodology used for the simulations, including the use of the Quantics Package and the Zagreb SH code for surface hopping calculations. The authors use different representations of the potential surfaces, including the linear vibronic coupling (LVC) model, to compare the performance of the methods. The results show that the DD-vMCG method provides accurate and reliable results, while the TSH method has issues with overcoherence and energy conservation. The study highlights the importance of benchmarking non-adiabatic quantum dynamics methods using a common set of systems and observables. The authors conclude that the DD-vMCG method is a promising approach for simulating non-adiabatic dynamics, and that further research is needed to improve the accuracy and efficiency of the method. The results also show that the choice of initial conditions and the representation of the potential surfaces significantly affect the results of the simulations. The study provides a comprehensive benchmark for future research in non-adiabatic quantum dynamics.This paper presents a benchmarking study of non-adiabatic quantum dynamics using the molecular Tully models. The authors validate the on-the-fly direct dynamics variational multi-configuration Gaussian (DD-vMCG) method using three molecular systems: ethene (IC1), DMABN (IC2), and fulvene (IC3). These systems were proposed by Ibele and Curchod as molecular versions of the Tully models used to test one-dimensional non-adiabatic behaviour. The study compares the performance of DD-vMCG with other methods, including Tully Surface Hopping (TSH) and Ab Initio Multiple Spawning (AIMS). The authors also test parametrised linear vibronic potential energy surfaces for each system and compare them to on-the-fly results. The three systems exhibit different deactivation pathways after excitation to their ππ* bright states. When comparing DD-vMCG to AIMS and TSH, the authors find crucial differences in some cases, which are explained by the classical nature and initial conditions of the TSH simulations. The study also explores the characterisation of conical intersections, which are classified as either peaked or sloped. The authors find that the conical intersection of DMABN is peaked, while the population in fulvene passes through a sloped conical intersection. The paper also discusses the methodology used for the simulations, including the use of the Quantics Package and the Zagreb SH code for surface hopping calculations. The authors use different representations of the potential surfaces, including the linear vibronic coupling (LVC) model, to compare the performance of the methods. The results show that the DD-vMCG method provides accurate and reliable results, while the TSH method has issues with overcoherence and energy conservation. The study highlights the importance of benchmarking non-adiabatic quantum dynamics methods using a common set of systems and observables. The authors conclude that the DD-vMCG method is a promising approach for simulating non-adiabatic dynamics, and that further research is needed to improve the accuracy and efficiency of the method. The results also show that the choice of initial conditions and the representation of the potential surfaces significantly affect the results of the simulations. The study provides a comprehensive benchmark for future research in non-adiabatic quantum dynamics.