<p>Many codecs with different error-correction approaches have been implemented for DNA data storage to date. However, no studies have systematically benchmarked codec implementations to establish their current state-of-the-art. Here, we use in silico and in vitro experiments to compare the performance of six representative codecs from literature. In isolation, these codecs can tolerate error rates up to 14% and a sequence loss of 65%. Under realistic conditions, we further establish that storage densities as high as 117 EB g⁻¹ are feasible using existing codecs and current synthesis and sequencing technologies. Verifying our results experimentally, we demonstrate data storage at 43 EB g⁻¹ using synthesis by material deposition and 13 EB g⁻¹ using electrochemical synthesis, employing existing codecs from literature. Besides closing in on the physical limits of DNA data storage, this study thus demonstrates the maturity of error-correction coding, defines its current state-of-the-art, and establishes best practices for codec benchmarking.</p>