NVIDIA’s Nemotron 3.5 Super VL is a large vision-language model supporting inputs of images, video, and text to generate textual outputs. It features a 120B-parameter Mixture-of-Experts backbone and vision encoder, accommodating inputs up to one million tokens. This model targets applications such as document intelligence, GUI understanding, and video question answering.
Post-training involved supervised fine-tuning on varied data types and reinforcement learning, enabling structured JSON outputs and configurable reasoning. Amazon SageMaker JumpStart provides two variants optimized for inference or customization. This rollout broadens accessible foundation models for developers deploying advanced multimodal AI solutions on AWS infrastructure.
Source & further reading
Nemotron 3.5 Super VL model now available on Amazon SageMaker JumpStart ↗Source published 2026-10-16. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.