This February 2026 benchmark update expands interval operations from three to eight and compares four libraries: polars-bio 0.24.0, PyRanges1 1.2.0, GenomicRanges 0.8.4, and bioframe 0.8.0. Tests use the AIList dataset, including the large 8-7 case with 307M overlap result pairs.
Key results at scale are clear: polars-bio is the fastest in 7 of 8 operations on the large dataset, while GenomicRanges wins nearest with a 1.63x advantage. Bioframe is consistently the slowest, trailing polars-bio by roughly 5x to 50x depending on the operation. For new operations (coverage, cluster, complement, merge, subtract), polars-bio leads across the board.
Threading matters too. On the large 8-7 dataset, most operations show 5-6x speedup at 8 threads, and overlap completes in 0.74 seconds at 8 threads. For full methodology, figures, and raw results, read the original post: https://biodatageeks.org/polars-bio/blog/2026/02/20/interval-operations-benchmark—update-february-2026/