NVIDIA Nemotron 3.5 Lightning Now Available in Amazon SageMaker JumpStart

NVIDIA Nemotron 3.5 Lightning has been introduced to Amazon SageMaker JumpStart, providing users with an open, high-performance foundation model designed for high-volume agentic workloads. This model leverages a hybrid Mixture-of-Experts (MoE) architecture with 30 billion total parameters, of which 3 billion are active during each forward pass, enabling efficient processing on supported GPUs.

The model is optimized for tasks that require continuous, agentic operation such as managing emails, financial document processing, cybersecurity alert enrichment, and customer service interactions. Its architecture supports up to 1 million tokens in context, facilitating long, multi-turn sessions without the need for frequent re-grounding.

NVIDIA reports that Nemotron 3.5 Lightning can deliver up to four times higher throughput and complete tasks approximately 30% faster compared to previous models, making it suitable for enterprise environments where speed and efficiency are critical. The model's open nature allows organizations to customize and deploy it across various platforms, including their own tools and workflows.

Deployment is straightforward through Amazon SageMaker Studio or via the Hugging Face model hub, with options for configuration using the SageMaker Python SDK. The model is suitable for a range of sectors, including personal assistants, financial services, cybersecurity, telecom, and retail.

Users should be aware that deploying the model will incur charges for the SageMaker endpoint while active, and it is recommended to delete endpoints when not in use to avoid ongoing costs. The model's architecture and performance specifications are detailed in the official SageMaker JumpStart documentation.