## This item has been replaced by the one which can be found at https://doi.org/10.7488/ds/2296. ## The cow rumen is a specialised organ adapted for the efficient breakdown of plant material into energy and nutrients, and it is largely the rumen microbiome that encodes the enzymes responsible. Many of these enzymes are of significant industrial interest. Despite this, rumen microbes are under-represented in public databases. Here we present 283 draft bacterial and archaeal genomes assembled directly from over 800 gigabases of rumen metagenomic sequence data and 43 samples, using both metagenomic binning and Hi-C-based Proximity-Guided Assembly. Comparative analysis with current publicly available genomes reveals that the majority of these represent previously unsequenced strains and species of bacteria and archaea. The genomes contain over 16,000 proteins predicted to be involved in carbohydrate metabolism, over 90% of which do not have a good match in public databases. Inclusion of the 283 genomes presented here improves metagenomic read classification by 2-3-fold, both in our data and in other publicly available rumen datasets. This release improves the coverage of rumen microbes in the public databases, and represents a highly valuable resource for biomass-degrading enzyme discovery and studies of the rumen microbiome.