Long series of temperature measurements in the winter Arctic stratosphere have been obtained from infrared Montgolfier (MIR) long-duration balloon flights in 1997, 1999, and 2000, from which the performance of a variety of meteorological analyses has been investigated. In this first paper of a series of two, the experimental setup and the measurements are described. First flown in the winter Arctic in 1997, the MIR platform appears well adapted to perform flights for several weeks in the vortex. Flights of 12 and 22 days (1997), 7 and 17 days (1999), and 3 and 18 days (2000) have been achieved along which altitude, pressure, and temperature have been sampled every 9–10 min. Comparisons between various independent sensors on the same balloon, and to Vaisala and Russian radiosondes, have made it possible to evaluate the performances of the instruments. The accuracy of the altitude/location of the Global Positioning System proved to be ±100 m but that of the pressure sensors only ±2 hPa. The most accurate method for deriving pressure appears to be the use of GPS altitude together with the ECMWF (European Centre for Medium-Range Weather Forecasts) geopotential height. Finally, the temperature was demonstrated to be measured with a precision of ±0.4 K and an average bias of less than ±0.5 K, but during nighttime and at an altitude below only 28 km.