We present a biological case study of codon bias with the probabilistic model checker PRISM with which we perform a quantitative analysis of expression speeds of genes. The variability in this setting concerns the matching of codons and anticodons. We distinguish between iso-acceptance (one codon matches exactly one anticodon) and wobble matching (multiple codons are matched by one anticodon). Our modeling confirms a recent result by Lavner and Kotlar, stating that at low and high expression speeds high codon bias prevails, whereas at moderate expression speeds low codon bias is more advantageous. Although the approach needs further validation, our preliminary investigation shows that probabilistic model checking with PRISM is promising in obtaining further insight in codon bias.