Nvidia Unveils Nemotron 3 Super with Enhanced Throughput
- Nvidia’s Nemotron 3 Super is a new AI model activating only 12.7 billion parameters per forward pass.
- The model delivers up to a remarkable 7.5 times more throughput than Qwen3.5-122B-A10B in agent workloads.
- Nemotron 3 Super supports context windows up to one million tokens without typical memory penalties.
- The model scores highly on benchmarks, achieving an impressive score of 91.64 on RULER long-context evaluation.
- It is fully open under the Nvidia Nemotron Open Model License with resources available on Hugging Face.
Nvidia’s latest AI model, Nemotron 3 Super, offers significant improvements in throughput and efficiency for agentic workloads by activating only a fraction of its parameters per pass, thus reducing costs and resource usage.
With its advanced architecture and open accessibility, the Nemotron model sets new standards in AI performance, particularly in handling extensive context windows efficiently.
(Source)
