Vicia faba (faba bean) is a nitrogen-fixing protein crop that, in contrast to soya bean, yields well in temperate climates. It is diploid but has a very large genome of ~13 Gb, and genomic resources are currently limited. To provide a high-quality reference gene set and expression atlas, we have carried out RNA-seq of multiple tissues from five faba bean accessions. For each tissue, an assembly was produced using the Trinity software package using both Illumina and PacBio reads. For each accession, the assemblies were then merged to create a more complete reference gene set. The assemblies were of high quality, with more than 95% complete BUSCOs [1] and read mapping rates of ~90%. The assemblies for each accession were then joined to create a comprehensive reference gene set along with an expression atlas describing gene expression patterns across accessions and tissues. These resources will now facilitate gene discovery and characterisation in faba bean and speed up identification of genes associated with important traits including pest and disease resistance and antinutrients such as vicine and convicine.