The University of Washington’s Institute for Protein Design (IPD) and Nvidia unveiled RFD4-Proteina, an open-source generative model aimed at improving protein design success rates and throughput. The work is built on Nvidia’s NVIDIA BioNeMo platform and is intended to make design automation faster while keeping the model accessible for broader research use. The developers said the model generates targeted protein designs and supports search-and-refine workflows using molecular inputs, constraints, and scoring objectives. In benchmarking, they reported five- to ten-fold higher design success rates and faster candidate generation compared with prior models. The system reportedly refines protein sequence and structure while co-evolving interacting molecule structures, enabling more integrated design iterations for binding, enzyme scaffolding, and DNA-binding tasks. The teams also positioned openness as a way to let researchers extend capabilities and accelerate lab testing. For drug discovery organizations, the launch signals continued movement toward open, foundation-like protein design tooling paired with scalable compute—an approach that can affect pipeline timelines if benchmarked results translate into higher hit rates for wet-lab programs.