NVIDIA open-sources BioNeMo inference runtime, cutting protein folding VRAM to one-third
AllThingsApx · x · 2026-09-12
- NVIDIA announced the public beta of BioNeMo Inference Runtime, an open PyTorch-native library that accelerates biomolecular inference on NVIDIA GPUs using specialized kernels and CUDA Graphs, supporting models like Boltz-2, OpenFold2, and Protenix v2, and scaling with Ray-powered GPU replicas.
- A third-party lab (@nanogenomic) benchmarked it against their production folding stack: faster with roughly one-third the VRAM per fold; the thread then checks whether outputs match the originals.
- Built with collaborators including apheris, SandboxAQ, and Xaira, it marks a major cost reduction for protein structure prediction infrastructure.
More from Infra
- EdgeRock-4B holds KV cache at 72 MiB across 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- RTX 5080 price jumps from $1658 to $2318 in ten days on Amazon — Frizzy-MacDrizzle · 2026-10-27
- ComfyUI on 12GB VRAM: measured peaks show offloading turns 24s into 11 minutes — Impossible-Mail-4619 · 2026-10-27
- Neon rebuilds its docs search entirely on its own stack: Lakebase Search, Functions, AI Gateway — matei_zaharia · 2026-10-27
- PyTorch details 2.12x faster on-device streaming ASR on Arm SME2 via ExecuTorch — PyTorch · 2026-10-27
- NVIDIA shows speculative decoding speeding up LLM inference on Jetson with Qwen3.5 9B — NVIDIA Developer · 2026-10-27