The world is reaching a point where the effectiveness of antibiotics could be completely compromised in the near future, if antimicrobial resistance (AMR) continues to spread globally. The intestinal microbiota of human and domestic animals is suspected to be the main reservoir of AMR bacteria and therefore of antibiotic resistance genes (ARGs) . Advances in DNA sequencing technologies and more specifically metagenomic approaches have shown that several ARGs are prevalent and shared between most gut microbiota , . Although nucleotide diversity is documented for well-known ARGs, the specific distribution of each ARG variant in gut resistomes is still completely unknown. This study aims to understand if ARGs described as identical in different microbiomes are actually the same variants. We adapted DESMAN , algorithm previously developed to reconstruct bacterial strain genomes from metagenomic data, to characterize ARG variants from a pool of ARGs. This tool was applied to ARGs detected in metagenomic cecal samples of chickens raised under two different conditions. Among 22 dominant ARGs analyzed, 15 have at least 2 stable variants and 7 (ant(6), blaTEM-1, erm(B), erm(F), tet(Q), tet(L) and tet(32)) have variants showing different proportions between cecal metagenomes depending on rearing condition and age of animals. Sequence comparison between ARG variants reconstructed by the DESMAN tool and those identified from bacterial isolates revealed a perfect match. These results attest to the reliability of this tool to reconstruct different variants of the same ARG directly from metagenomic data and to infer their relative proportions in different samples. It opens the way for further analyze the relative abundance of ARG variants in metagenomic datasets and finally deciphering their transmissions between microbiomes of different hosts.