This research evaluated the performance of Land Surface Models (LSMs) in simulating droughts, examining Land-Hist offline simulations from the Land, Surface, Snow and Soil Moisture Intercomparison Project (LS3MIP). It is well known that LSMs possess uncertainties and biases due to oversimplifications or the absence of certain physical processes (e.g., groundwater interactions and lateral connectivity). Therefore, the objective of this research was to identify the strengths and weaknesses of various LSMs and how this relates to the performances in simulating soil moisture droughts. To address this objective, eight LSMs were evaluated: CESM2, CMCC-ESM2, E3SM-1-1, EC-Earth3-Veg, HadGEM3-GC31-LL, IPSL-CM6A-LR, MIROC6, and UKESM1-0-LL. Two reference evaporation data sets (DOLCE V3 and an ensemble of FLUXCOM-RS, BESS and PML) and a reference soil moisture data set (SoMo.ml) were utilized for the evaluation. After a global analysis on the LSM evaporation characteristics, six climate diverse study areas were selected for further investigation. A long-term analysis was performed by examining the water balance and implementing the LSMs into the Budyko Framework. Subsequently, soil moisture deficits were calculated for the driest periods in time, and the resulting accumulated deficits were compared with the reference evaporation data. The timing and progression of the deficits were evaluated utilizing the reference soil moisture data. Finally, the sensitivity of the model was evaluated by examining the response of evaporation anomalies to precipitation anomalies and comparing this with the reference evaporation data. The results showed that there was a large spread in output and performance among the LSMs across all parts of the evaluation. The greatest contrasts among the LSMs were found in the dry to wet transition zones within the tropics. In this latitudinal range, the worst performing LSMs overestimated the accumulation of soil moisture deficits and the severity of droughts, while the opposite was found for the extratropical regions. Additionally, the models showed, in general, to be overly sensitive to precipitation anomalies. When ranking the implemented model bases in the LSMs based on their performance during droughts, the findings showed that the Community Land Model (implemented in CMCC-ESM2, E3SM-1-1 and CESM2) was predominantly the best performing, followed by ORCHIDEE (IPSL-CM6A-LR) and HTESSEL (EC-Earth3-Veg). MATSIRO (MIROC6) and JULES (HadGEM3-GC31-LL and UKESM1-0-LL) were the least performing model bases. From a hydrological perspective, the findings of this research could be linked to some known limitations of LSMs. Oversimplified soil and vegetation dynamics could contribute to the LSMs being overly sensitive to precipitation anomalies while the contrasts between the tropical and extratropical regions could be attributed to the representation of the soil moisture-evaporation coupling, which plays a greater role in the tropical study areas. Ultimately, this research could contribute to LS3MIP and the Land Surface Modeling community, as the results highlight the strengths and weaknesses of the LSMs in simulating soil moisture droughts. From there, this research could contribute to improving LSMs, understanding drought mechanisms, and addressing climate change impacts, especially in drought-prone regions.