Quantifying gene expression levels is an important research tool to understand biological systems. Reverse transcription-quantitative real-time PCR (RT-qPCR) is the preferred method for targeted gene expression measurements because of its sensitivity and reproducibility. However, normalization, necessary to correct for sample input and reverse transcriptase efficiency, is a crucial step to obtain reliable RT-qPCR results. Stably expressed genes (i.e. genes whose expression is not affected by the treatment or developmental stage under study) are indispensable for accurate normalization of RT-qPCR experiments. Lack of accurate normalization could affect the results and may lead to false conclusions. Since transcriptomes of seeds are different from other plant tissues, we aimed to identify reference genes specifically for RT-qPCR analyses in seeds of two important seed model species, i.e. Arabidopsis and tomato. We mined Arabidopsis seed microarray data to identify stably expressed genes and analyzed these together with putative reference genes from other sources. In total, the expression stability of 24 putative reference genes was validated by RT-qPCR in Arabidopsis seed samples. For tomato, we lacked transcriptome data sets of seeds and therefore we tested the tomato homologs of the reference genes found for Arabidopsis seeds. In conclusion, we identified 14 Arabidopsis and nine tomato reference genes. This provides a valuable resource for accurate normalization of gene expression experiments in seed research for two important seed model species.