Thermal recovery in boreholes cooled by circulation of drilling mud has been modeled for estimating formation temperature and thermal conductivity. Coupled with a finite-element simulation of heat conduction, inverse modeling for the desired parameters starts with a genetic algorithm that feeds initial estimates of model parameters to an iterative quasi-linear inversion scheme. In addition to using the rms misfit between the computed and observed borehole temperatures, the results are assessed by comparing or constraining the model formation temperature with a value obtained conventionally from an asymptotic temperature–time relation for a steady line source. The model conductivity is further evaluated for equality with a conductivity value, which is estimated through simulation of heat exchange between the formation and circulating mud. Test results on synthetic data and two sets of highly noisy borehole data from Lake Baikal in Russia indicate that the two equality criteria in temperature and conductivity are achievable. Multiple runs of GA-IM are used to find mean parameter values and their uncertainties. The resultant model conductivity values are consistent with those measured in cores with a needle-probe method.