The open MoE model activates 3B of 30B parameters per forward pass for up to 4x throughput and 30% faster task completion.

NVIDIA Nemotron 3.5 Lightning is now available in Amazon SageMaker JumpStart. Operators can deploy the open model without configuring serving infrastructure. The release targets high-volume agentic workloads that demand fast, specialized model execution.

NVIDIA built Lightning by distilling its frontier Nemotron 3 Ultra model and developed it with the Nemotron Coalition. The hybrid Mixture-of-Experts architecture activates 3B of 30B parameters per forward pass. NVIDIA also trained the model on open datasets and released it as an open model, so builders own any custom weights.

The small active-parameter count lets Lightning run on a single supported GPU, so operators can reserve frontier-scale models for planning and orchestration. Builders can route routine steps like classification and data extraction to Lightning, cutting latency and cost. The system-of-models approach suits always-on agents that gather context and act continuously.

NVIDIA positions Lightning as the fastest open model in its class for always-on agents, and SageMaker JumpStart removes the serving setup burden. Builders should test it against their agent harnesses and measure throughput on high-volume tasks. Watch for more open models with the same distillation approach to appear on managed platforms.

What matters

• NVIDIA Nemotron 3.5 Lightning is now deployable from Amazon SageMaker JumpStart.
• Builders can run high-volume agentic workloads on a single GPU with up to 4x throughput.
• Watch for enterprise adoption of the open Nemotron 3 Ultra distillate across agent harnesses.

Why it matters

Watch for enterprise adoption of the open Nemotron 3 Ultra distillate across agent harnesses.

This GenAI News article was prepared in original wording using reporting and materials published by AWS Machine Learning Blog. Source reference: https://aws.amazon.com/blogs/machine-learning/nvidia-nemotron-3-5-lightning-now-available-in-amazon-sagemaker-jumpstart/.

Drafted by the GenAI News review pipeline.