Oryza is remarkable genus - with two domesticated (i.e. Asian and African rice) and 25 diploid and tetraploid wild species, 11 extant genome types, and a ~3.4-fold genome size variation - that possesses a virtually untapped reservoir of genes that can be used for crop improvement. Here we unveil and interrogate 11 new chromosome-level assemblies of nine tetraploid and two diploid wild Oryza species in the context of ~15 million years of evolution of the genus. We show that the core Oryza (sub)genome across all genome types is only ~200 Mb and largely syntenic, while the remaining nuclear fractions, spanning ~80-600 Mb, are intermingled, extremely plastic and rapidly evolving. For the halophyte O. coarctata, we show that - despite the detection of gene fractionation in the subgenomes - homoeologous genes are expressed at higher levels in one subgenome over the other in a mosaic form, thereby showing subgenome equivalence. The integration of these 11 new ultra-high quality reference genomes with our previously published genome data sets provide a nearly complete picture of the consequences of natural and artificial selection across the evolutionary history of Oryza. This in turn opens the door to unlock their genetic potential for future crop improvement and neodomestication.