The University of Washington’s Institute for Protein Design (IPD) and Nvidia released an open-source protein design AI model aimed at improving success rates and speed for generating proteins with specific functions. The model, RFD4-Proteina, is built on Nvidia’s NVIDIA BioNeMo framework and uses generative protein design with search-based refinement. Developers said it achieved five-fold to 10-fold higher design success rates while producing protein candidates five to 10 times faster than prior models in benchmark testing. In lab bench evaluations, the system reported improved computational design outcomes across protein binding, small-molecule binding, enzyme scaffolding, and DNA binding. Remaining open-source is positioned as a way to let researchers extend and evaluate the model across diverse drug-discovery workflows. For biotech, open tools can accelerate reproducible protein-candidate generation and downstream screening, especially for targets where conventional binder engineering struggles with speed and iterate-to-lead costs.