Deploy NVIDIA Nemotron 3.5 Super VL on AWS EKS and SageMaker HyperPod EKS

It supports multimodal inputs including text, image, document, and video with reasoning and tool use. The hybrid Mamba-2 and MoE architecture activates 12B parameters per token, enabling efficient long-context inference. Performance reaches 63.44% on Hermes Bench, 87.05% on LangChain agent tasks, and 80.8% on CharXiv chart reasoning. Deployment uses NVIDIA Dynamo and vLLM across standard EKS, EKS Auto Mode, and HyperPod EKS with GPU disaggregation and NCCL validation.

Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.