Nvidia Nemotron 3 Ultra Tops Open U.S. AI Models With 48-Point Score
Updated
Updated · THE DECODER · Jun 1
Nvidia Nemotron 3 Ultra Tops Open U.S. AI Models With 48-Point Score
6 articles · Updated · THE DECODER · Jun 1
Artificial Analysis ranked Nvidia’s Nemotron 3 Ultra the strongest open U.S. model, giving it 48 points—above Gemma 4 31B at 39, Nemotron 3 Super at 36 and gpt-oss-120b at 33.
The model uses roughly 550 billion total parameters with about 55 billion active at a time, and Artificial Analysis said it delivers more than 300 tokens per second on DeepInfra.
That still leaves Nvidia behind China’s top open model: Kimi K2.6 scored 54 points, while the leading closed model, Opus 4.8, reached 61.
Nvidia said Nemotron 3 Ultra will be released on June 4 through Hugging Face, OpenRouter and other platforms, widening access to its highest-ranked open U.S. system.
Can Nemotron's raw speed overcome the cost advantage of Chinese rivals in the global AI adoption race?
Is Nvidia's new open model a strategic counter to China's growing dominance in the open-source AI race?
Nemotron 3 Ultra Launch: Nvidia’s 550B-Parameter Open-Weight AI Sets New Standard for Speed and Cost-Efficiency
Overview
Nvidia unveiled its flagship AI model, Nemotron 3 Ultra, at Computex 2026 in Taipei, marking a major step into the open-weight AI domain. This model, with 500-550 billion parameters, stands at the top of the Nemotron 3 family, which also includes the Super and Nano variants. Nvidia’s goal is to empower developers and meet the rising demand for advanced, cost-effective AI solutions. By introducing Nemotron 3 Ultra, Nvidia aims to provide powerful tools for innovation, making AI more accessible and efficient for a wide range of applications.