The GPS (Global Positioning System) radio occultation experiment onboard the German CHAMP (CHAllenging Minisatellite Payload) satellite was successfully started in February 2001. By the end of July 2001 more than 11,000 globally distributed vertical profiles of atmospheric temperature were derived. Operational data processing is based on the double difference method to correct for satellite clock errors. This technique requires a continuous data stream from a global network of GPS ground stations. Termination of the GPS Selective Availability mode on May 2, 2000 reduced the GPS clock errors by orders of magnitude, thereby rendering space‐based single differencing feasible. First results using this single difference technique without direct use of GPS ground station data are presented. The comparison of two data sets consisting of 436 occultations, one processed with space‐based single differencing, the other with classical double differencing, yields no discernible temperature bias below 30 km altitude, a standard deviation of <0.6 K below 20 km and <1.2 K below 30 km. The comparison of both data sets with corresponding meteorological analyses shows nearly identical results with differences in mean and standard deviation of <0.15 K up to 30 km.