International audienceIn the last decades, density functional theory has become unavoidable in theoretical studies of organometallic chemistry. Most of the recent functionals contain many parameters that are adjusted using carefully chosen reaction sets. However, these sets only contain a few entries involving late transition metal reaction, so that choosing a functional for such a study is difficult. In this work, the theoretical description of the oxidative addition of Pd(PH3)2 to 2-iodo-allyl-aniline was chosen as a representative reaction of palladium. The competitive binding of the palladium to the alkene or the nitrogen atom was used to assess the accuracy of ab initio methods (MP2, MP3, MP2.5, SCS-MP2, SCS-MP3) and 56 functionals ranging from local density approximation to the costly double-hybrid approaches (such as B2PLYP), against a CCSD(T)/CBS reference value. Model systems [(PH3)2ClPd(NH3)]+ and [(PH3)2ClPd(H2C=CH2)]+ were first considered: all functionals correctly predict that the azane complex is the most stable. However, some functionals overestimate its stability compared to the alkene complex. This is amplified in the 2-iodo-allyl-aniline study: SCS-MP3, B2PLYP as well as BP86, most of the meta-GGA (generalized gradient approximation), hybrid GGAs and hybrid meta-GGAs are predicting that oxidative addition proceeds directly. On the contrary, many functionals, among which B3LYP, M06-2X and most range-separated methods, wrongly predict that palladium first binds to the nitrogen atom before proceeding to the olefin insertion. Resorting to these functionals to study inorganic reactions with palladium might thus result in predicting wrong mechanisms