NVIDIA Expands Open Model Families for Physical AI and Healthcare — Nemotron 3 Nano Available Now

NVIDIA Newsroom·medium signal

Alongside Nemotron 3 Super, NVIDIA announced Nemotron 3 Nano — a 30B total / 3B active parameter model available now for DGX Spark, H100, and B200 that delivers 4x higher throughput than Nemotron 2 Nano. The full family (Nano available, Super/Ultra coming H1 2026) targets agentic, physical AI, and healthcare verticals, with Ultra aimed at the highest-complexity reasoning tasks. The strategy positions NVIDIA to own the open model stack for enterprise verticals where proprietary models face regulatory or latency constraints.