We developed and implemented in the 3D air quality model CHIMERE the formation of several key anthropogenic aerosol markers including one primary anthropogenic marker (levoglucosan) and 4 secondary anthropogenic markers (nitrophenols, nitroguaiacols, methylnitrocatechols and phthalic acid). Modelled concentrations have been compared to measurements performed at 12 locations in France for levoglucosan in winter 2014–15, and at a sub-urban station in the Paris region over the whole year 2015 for secondary molecular markers. While a good estimation of levoglucosan concentrations by the model has been obtained for a few sites, a strong underestimation was simulated for most of the stations especially for western locations due to a probable underestimation of residential wood burning emissions. The simulated ratio between wood burning organic matter and particulate phase levoglucosan is constant only at high OM values (>10 μg m−3) indicating that using marker contribution ratio may be valid only under certain conditions. Concentrations of secondary markers were well reproduced by the model for nitrophenols and nitroguaiacols but were underestimated for methylnitrocatechols and phthalic acid highlighting missing formation pathways and/or precursor emissions. By comparing modelled to measured Gas/Particle Partitioning (GPP) of markers, the simulated partitioning of Semi-Volatile Organic Compounds (SVOCs) was evaluated. Except for nitroguaiacols and nitrophenols when ideality was assumed, the GPP for all the markers was underestimated and mainly driven by the hydrophilic partitioning. SVOCs GPP, and more generally of all SVOC contributing to the formation of SOA, could therefore be significantly underestimated by air quality models, especially when only the partitioning on the organic phase is considered. Our results show that marker modelling can give insights on some processes (such as precursor emissions or missing mechanisms) involved in SOA formation and could prove especially useful to evaluate the GPP in 3D air quality models.