Hi Everyone, We have been benchmarking common computational biology workloads to better understand performance and cost tradeoffs. Sharing a few highlights: • AlphaFold 3 → 33 to 135 minutes runtime depending on system size | up to 67 percent lower cost | two stage CPU+GPU strategy improves both pipeline and inference • Boltz 1 → 16 to 21 minutes per run | as low as $0.10 per prediction | accuracy close to AlphaFold 3 across proteins, RNA, and DNA • OpenMM → as low as $6.59 per microsecond | 3359 ns per day | flexible optimization for speed, cost, or balance • GROMACS → benchmarked across 20k to 1M atoms | up to 1139 ns per day | $28.64 per microsecond (34.9 ns per dollar) • MegaDOCK → 1,000 antibody antigen dockings completed in about two hours for $3.09 | scales to millions of dockings in hours • AutoDock Vina GPU → 30,000 ligands screened at under $1 per 10,000 ligands | QuickVina2 variant ran 1,000 ligands in 15 minutes The intent is to give practical reference points for anyone running these workloads or designing pipelines around them. If you are facing challenges with Nextflow or scaling computational workloads, feel free to DM me.