A semiempirical thermosphere model of Mars has been constructed based on the DTM algorithm that was developed for the terrestrial thermosphere. DTM-Mars reproduces the observed densities with a 1-σ model uncertainty of approximately 35% outside dust season. During dust storms the uncertainty increases by approximately a factor two. The data used in this study have been derived from the Mars Global Surveyor accelerometer observations as well as from analysis of precise orbit determination results. The temperature modeling is based on the Mars-GRAM 2000 [ Justus and James, 2000 ] and the MTGCM [ Bougher et al., 2000 ] models, while the CO₂, O, and He constituent modeling has been achieved through the adjustment of model coefficients to density data. DTM-Mars is preliminary because the geographical and solar activity coverage of the assimilated density data currently available is not representative of all atmospheric conditions. For example, the effect of dust storms on the thermospheric density is only represented by a temperature increase on a planetary scale, because the data did not allow the determination of a semidiurnal variation caused by local effects in the lower atmosphere. The model accuracy will be improved in the future by assimilating density data taken under different atmospheric conditions, for example, from the Mars Odyssey mission, once these become available.