NVIDIA's RFD4-Proteina lands on HuggingFace, outdesigning rivals 5.8x per GPU-hour
AllThingsApx · x · 2026-10-21
NVIDIA and UW's Institute for Protein Design have open-sourced RFD4-Proteina on HuggingFace. The model has an LLM-like "thinking" setting: at high thinking it produces 76.40 usable designs per GPU-hour on a 24-binding-target benchmark, versus 13.2 for the next best method.
It jointly designs protein sequences and all-atom structures conditioned on molecular context — binding targets, enzyme motifs, ligands, or selected structural features. A demo space, weights and blog post are available.
Related event: NVIDIA and Baker Lab Open-Source RFD4-Proteina Protein Design Model(35 posts)→
More from Research
- Paper asks 'How long is the coast of a neuron?' — resolution shapes neural computation — neurovium · 2026-10-21
- Yoav Goldberg on ICLR reviewing: question-title papers must actually answer the question — yoavgo · 2026-10-21
- Surge AI's Chartography and GDP.pdf top Papers with Code trending benchmarks — NielsRogge · 2026-10-21
- OpenAI scales thousands of AI agents to find new leads in biological data — truecakesnake · 2026-10-21
- EAU: Co-evolving user simulator, environment, and RL setup for training LLM agents — UAlberta · 2026-10-21
- Meta's Self-Credit: Dense Rewards from Sparse RLVR via Intra-group Rescaling — meta · 2026-10-21