A model for understanding documents and screens
Following AWS's November 9 announcement, Nemotron 3.5 Super VL can be deployed through SageMaker JumpStart. It has 120 billion parameters, with 12 billion active per token.
The model supports text, image and video inputs, with a context limit of 1 million tokens. Target workloads include document review, screen-based agents and questions about video.
Two weight formats
JumpStart offers BF16 and NVFP4 options. The NVFP4 version is optimized for Blackwell GPUs. Server costs, hardware and request concurrency depend on the deployment; access to open weights does not mean free cloud inference.
