OSS
NVIDIA Announces Nemotron 3 Ultra at Computex — 550B Parameter Open-Weights Model With 55B Active via MoE
Jensen Huang unveiled Nemotron 3 Ultra at Computex 2026 today, a 550B-parameter mixture-of-experts model with 55B active parameters that tops US open-weights rankings with an Intelligence Index of 48 and delivers 300+ tokens/second. The model sits atop a three-tier family (Nano, Super 120B, Ultra 550B) and will be available on Hugging Face, ModelScope, OpenRouter, and as an NVIDIA NIM microservice. Designed specifically for advanced reasoning, planning, and agentic workflows — making it the most capable US open-weights model for autonomous multi-step tasks.
↳ Follow the thread