NVIDIA Nemotron 3 Nano 30B Model Available in Amazon SageMaker JumpStart
NVIDIA's Nemotron 3 Nano 30B language model is now accessible through Amazon SageMaker JumpStart. The model aims to enhance generative AI application development and offer scalable solutions for complex tasks.
Amazon Web Services (AWS) has made NVIDIA's Nemotron 3 Nano 30B language model available through its Amazon SageMaker JumpStart service. This open-weight model, featuring 3 billion active parameters, is designed to accelerate the development of generative AI applications and provide solutions for demanding tasks.
The Nemotron 3 Nano is a hybrid "mixture of experts" (MoE) model combining Transformer and Mamba architectures. The model reportedly offers high computational efficiency and accuracy, particularly in coding, reasoning, and mathematical tasks. Its open nature allows for customization and deployment within enterprises' own environments, aligning with their privacy and security requirements.
Available via SageMaker JumpStart, the model simplifies deployment without complex model management. AWS aims to provide developers with tools to speed up innovation and create business value using generative AI.
The model's performance has been evaluated in benchmarks such as SWE Bench Verified, GPQA Diamond, and AIME 2025, where it has achieved strong results, especially in coding and scientific reasoning tasks.