github nanoporetech/dorado v2.1.0

13 hours ago

[2.1.0] (13 July 2026)

This version of Dorado introduces performance improvements for v6.0.0 HAC models on Blackwell GPUs. Performance of v5.0.0+ SUP models has also been improved, especially for short reads. This release also contains a collection of feature improvements and bug fixes for Dorado smallvar, including initial gVCF support.

  • 9829da3 - Update to Koi 0.7.3 with improved Blackwell performance for v6.0.0 HAC models.
  • 0917d62 - Add support for variable chunk size basecalling for v5.0.0 onwards SUP models, improving performance for short reads.
  • b779ea1 - Improve the CPU efficiency of read splitting.
  • 4fa5164 - Support gVCF output from Dorado smallvar via the --gvcf option.
  • 1946d7a - Correctly report variants in Dorado smallvar when the reference is lower case.
  • c5a3cde - Sort and deduplicate processed_regions.bed file in Dorado smallvar.
  • 44442d1 - Add --hemizygous-regions option to Dorado smallvar, to specify regions in which haploid variant calls are expected.
  • 9f11663 - Remove duplicated fileformat header line from VCF output.
  • bd56aac - Bug fix for overlapping simple variants by removing redundant allele coverage check in Dorado smallvar.
  • b72ad62 - Fix output of overlapping variants due to trimming of processed regions before merging.
  • 36e332a - Add --MD, --cs and --rmq to the list of supported minimap options.
  • 7f72ddc - No longer create a .fai index file on disk alongside alignment references.
  • ab8217e - Fix sp:i tag to contain the correct split points in all cases when a read has been split multiple times.
  • ae09deb - Add support for single-ended dual barcode arrangements (two barcodes at one end of a read) and arrangements with split inter-barcode flanking regions.
  • 9ba24c5 - Port the Dorado correct Herro model from TorchScript to Libtorch.
  • 765ea19 - Remove use of TorchScript from Dorado.
  • 7cbf1c3 - Update to minimap2 v2.31.
  • 8678916 - Pin POD5 version to 0.3.39.
  • 7d36a47 - Fix batch size benchmarking while running LSTM models with variable chunk sizes.

Don't miss a new dorado release

NewReleases is sending notifications on new releases.