Background: Rice blast is one of the most damaging disease of rice. The use of resistant cultivars is the only practical way to control the disease in developing countries where most farmers cannot afford fungicides. However resistance often breaks down. Genome wide association studies (GWAS) allow high resolution exploration of rice genetic diversity for quantitative and qualitative resistance alleles that can be combined in breeding programs to achieve durability. We undertook a GWAS of resistance to rice blast using a tropical japonica panel of 150 accessions genotyped with 10,937 markers and an indica panel of 190 accessions genotyped with 14,187 markers. Results: The contrasted distribution of blast disease scores between the indica and tropical japonica groups observed in the field suggest a higher level of quantitative resistance in the japonica panel than in the indica panel. In the japonica panel, two different loci significantly associated with blast resistance were identified in two experimental sites. The first, detected by seven SNP markers located on chromosome 1, colocalized with a cluster of four NBS-LRR including the two cloned resistance genes Pi37 and Pish/Pi35. The second is located on chromosome 12 and is associated with partial resistance to blast. In the indica panel, we identified only one locus associated with blast resistance. The three markers significantly detected at this locus were located on chromosome 8 in the 240 kb region carrying Pi33, which encompasses a cluster of three nucleotide binding site-leucine-rich repeat (NBS-LRRs) and six LRR-kinases in the Nipponbare sequence. Within this region, there is an insertion in the IR64 sequence compared to the Nipponbare sequence which also contains resistance gene analogs. Pi33 may belong to this insertion. The analysis of haplotype diversity in the target region revealed two distinct haplotypes, both associated with Pi33 resistance. Conclusions: It was possible to identify three chromosoma