New results from the Nextflow Frontier team: on 22 BioML-bench tasks, Nextflow Frontier beat 96% of the prior methods on each leaderboard on average, against 75% for OpenAI’s Rosalind and 66% for Anthropic’s Claude Science. Rosalind and Claude Science stopped at the first answer that worked. Nextflow Frontier kept looking, with a median of 506 scored attempts per task.
The part pipeline writers will care about: every idea Nextflow Frontier tested was written as a Nextflow pipeline, with each part of the idea as its own process, and run at scale on Seqera’s platform.
Blog: Nextflow Frontier, Rosalind and Claude Science on BioML-bench — Nextflow Frontier