Nvidia Unveils 550B-Parameter Nemotron 3 Ultra as China Still Leads Open-Weight AI
Updated
Updated · Decrypt · Jun 1
Nvidia Unveils 550B-Parameter Nemotron 3 Ultra as China Still Leads Open-Weight AI
3 articles · Updated · Decrypt · Jun 1
Computex in Taipei marked Nvidia’s launch of Nemotron 3 Ultra, a 550-billion-parameter open-weight model with 55 billion active parameters that the company says will ship June 4.
Artificial Analysis scored the model at 48 on its intelligence index—12 points above Nemotron 3 Super and the highest among U.S. open-weight models—but still below Moonshot AI’s Kimi K2.6 at 54.
Over 300 tokens per second on a pre-release DeepInfra endpoint gives Ultra a major speed edge, with Nvidia claiming 3x-6x faster inference and 30% lower costs than comparable open-weight rivals.
A 1-million-token context window, mixture-of-experts design and public weights position the model for enterprise agents, though running it directly remains datacenter-scale.
The release is Nvidia’s clearest move yet in a disclosed $26 billion push to counter Chinese open-model gains, with Nemotron 4 already in development through an eight-lab coalition.
Is Nemotron 3 Ultra a gift to open-source AI, or a Trojan horse to deepen reliance on NVIDIA's proprietary hardware?
With China's Kimi K2.6 still ahead, can NVIDIA's $26 billion plan truly reverse the tide in the global AI race?
Nvidia Nemotron 3 Ultra: The 1-Million-Token Open-Weight AI Champion Challenging China’s Global Dominance
Overview
Nvidia has recently unveiled Nemotron 3 Ultra, a major leap in open-weight AI models and a result of the company’s focused AI investments. Set to launch in the first half of 2026, Nemotron 3 Ultra stands out for its high intelligence scores, fast output speed, and a groundbreaking 1-million-token context window. This large context window enables the model to deliver context-rich responses by seamlessly integrating data, past conversations, and code examples, supporting more coherent multi-agent workflows. Nemotron 3 Ultra’s advanced architecture and performance place it in the most attractive quadrant of industry analysis, highlighting Nvidia’s leadership in the evolving AI landscape.