SNP genotyping chip is an important tool for molecular breeding. High density SNP chip often has some problems such as marker redundancy, high price, and poor target, limiting its use in the normalization and scale of molecular breeding. Here we developed a low density breeding chip. The panel of 6,056 markers were assembled using three resources, consisting of: (1) 2,080 key SNPs of 18.2Mb SNPs which were identified from 10X sequencing of 37 maize inbred lines; (2) 3,390 qualified markers with low missing rate, high polymorphism and Conversion Type being as Poly High Resolution within 55K SNP array; (3) 586 markers that were selected from HapMap3. Genotyping by target sequencing (GBTS) technology was used to detect the markers. Through the verification of natural population, bi-parent population and multiparent advanced generation inter-cross (MAGIC) population, the average capture rate of the breeding chip was 70.6%. The original design markers were detected ranged from 4,773-5,963, and 57.6% and 88.6% markers were found by applying minor allele frequency (MAF) > 0.4 and polymorphic information content (PIC) > 0.4, respectively. We evaluated 226 inbred lines with this breeding chip. Within this collection two groups (temperate vs. tropical) had been classified by principal component analysis (PCA), and six known groups (Reid, Lancaster, PB, LRC, SPT and tropical) were proposed using cluster analysis. Structure analysis has not revealed the best K value. The mean genetic distance within and among groups were 0.394 and 0.472, respectively. The genetic distance within PB group was the smallest (0.316), and the genetic distance within tropical group was the largest (0.424). The largest genetic distance (0.493) was observed between group Reid and tropical. The genetic differentiation coefficient (FST) among the groups indicated that the FST of PB group was larger than that of other groups.