On October 21, 2026, NVIDIA announced the expansion of its open model product line, introducing the Nemotron Retriever retrieval model, Kumo structured enterprise data model, Cosmos 3.5 world model, Alpamayo 2 Edge autonomous driving model, BioNeMo protein design model, and Ising quantum computing model. Among them, the BioNeMo RFD4-Proteina achieved a design generation speed approximately 13 times faster than RFdiffusion3 in benchmark tests covering 200-amino-acid proteins.
