Grain number per spike (GNPS) is a major factor in wheat yield breeding. The development of high GNPS germplasm is widely emphasized in wheat-yield breeding. This paper reported two high GNPS wheat germplasm lines, Pubing 3228 and Pubing 3504, which had a stable and wide adaptability to different ecological regions. By exploring a nested cross design with reciprocals using Pubing 3228 or Pubing 3504 as a common parent and investigating the GNPS phenotypes of F 1 hybrids in 2007-2008 and F 2 populations in 2008-2009 of different cross combinations, the narrow-sense GNPS heritability was up to 49.58 and 52.23%, respectively. Genetic model analysis predictions suggested that GNPS in Pubing 3228 and Pubing 3504 was mainly controlled by additive genetic effects. Correlation analysis results between GNPS and 1 000- kernel weight (TKW) of F 2 populations showed that TKW was not influenced with the increase of GNPS. The good coordination among three yield components of spike number per plant (SNPP), GNPS, and TKW in the F 2 segregating population implied that selection of good candidate individuals in breeding programs would be relatively straightforward. Overall, our results indicated that Pubing 3228 and Pubing 3504 are two potential germplasm lines for yield improvement of GNPS in pedigree selection of wheat breeding. The development of high GNPS germplasm is greatly emphasized in wheat-yield breeding. This paper reports two high GNPS wheat germplasm lines, Pubing 3228 and Pubing 3504, which had a stable and wide adaptability to different ecological regions. By exploring a nested cross design with reciprocals using Pubing 3228 or Pubing 3504 as a common parent and investigating the GNPS phenotypes of F1 hybrids in 2007-2008 and F2 population in 2008-2009 of different cross combination, the narrow-sense GNPS heritability was up to 49.58 and 52.23%, respectively. Genetic model analysis predictions suggested that GNPS in Pubing 3228 and Pubing 3504 was mainly controlled by additive genetic effects. Correlation analysis results between GNPS and 1000-kernel weight (TKW) of F2 populations showed that TKW was not influenced with the increase of GNPS. The good coordination among three yield components of spike number per plant ( SNPP), GNPS, and TKW in the F2 segregating population implied that the selection of good candidate individuals in breeding programs would be relatively straightforward. Overall, our results indicated that Pubing 3228 and Pubing 3504 are two potential germplasm lines for yield improvement of GNPS in pedigree selection of wheat breeding.