AWS has added NVIDIA’s Nemotron 3.5 Super VL to Amazon SageMaker JumpStart, its catalog for deploying foundation models. The model takes images, video and text as input and generates text, with uses including document analysis, video question answering and understanding graphical interfaces.
It supports context lengths of up to 1 million tokens. JumpStart offers an NVFP4 variant for inference on specified G7 instances and BF16 reference weights on an ml.p5en.48xlarge instance for customization; AWS says the latter supports fine-tuning and other adaptation methods.
Why it matters
AWS customers building document or video applications can deploy the model through SageMaker JumpStart, while those needing domain-specific behavior can customize the BF16 weights; the source gives no pricing or performance comparisons.
