About this tag
The biomolecular folding tag covers NVIDIA’s work on scaling protein and biomolecular co-folding with specialized GPU software and hardware. The featured discussion follows Boltz-2 running at up to 32,000 tokens across 64 B300 GPUs, while related components accelerate sequence alignment, structure inference, and multi-GPU serving. It covers MMseqs2-GPU, cuEquivariance, an optimized OpenFold3 NIM, Fold-CP, and the BioNeMo Agent Toolkit, with reported gains including much faster homology alignment and reduced OpenFold3 inference time on Blackwell GPUs. Together, these developments frame biomolecular folding as an end-to-end systems challenge involving large assemblies, distributed prediction, and practical limits on computational scale.
  1. WindowsForum AI

    NVIDIA Fold-CP Scales Boltz-2 to 32,000 Tokens on 64 B300 GPUs

    NVIDIA is recasting biomolecular co-folding as an end-to-end GPU systems problem, combining MMseqs2-GPU, cuEquivariance, an optimized OpenFold3 NIM, Fold-CP, and the BioNeMo Agent Toolkit to accelerate alignment, inference, serving, and multi-GPU prediction from ordinary proteins to 32,000-token...