Bacterial plasmids substantially contribute to the rapid spread of antibiotic resistance, which is a crisis in healthcare today. Coevolution of plasmids and their hosts promotes this spread of resistance by ameliorating the cost of plasmid carriage. However, our knowledge of plasmid–bacteria coevolution is solely based on studies done in well-mixed liquid cultures, even though biofilms represent the main way of bacterial life on Earth and are responsible for most infections. The spatial structure and the heterogeneity provided by biofilms are known to lead to increased genetic diversity as compared with well-mixed liquids. Therefore, we expect that growth in this complex environment could affect the evolutionary trajectories of plasmid–host dyads. We experimentally evolved Shewanella oneidensis MR-1 with plasmid pBP136Gm in biofilms and chemostats and sequenced the genomes of clones and populations. Biofilm populations not only maintained a higher diversity of mutations than chemostat populations but contained a few clones with markedly more persistent plasmids that evolved via multiple distinct trajectories. These included the acquisition of a putative toxin–antitoxin transposon by the plasmid and chromosomal mutations. Some of these genetic changes resulted in loss of plasmid transferability or decrease in plasmid cost. Growth in chemostats led to a higher proportion of variants with decreased plasmid persistence, a phenomenon not detected in biofilms. We suggest that the presence of more stable plasmid–host dyads in biofilms reflects higher genetic diversity and possibly unknown selection pressures. Overall, this study underscores the importance of the mode of growth in the evolution of antibiotic-resistant bacteria.