Not established from the available sources.
NVIDIA · MODEL RELEASE TRACKER
NVIDIA Nemotron 3.5 Super VL
Nemotron 3.5 Super VL is a vision-language model in the Nemotron 3.5 family that accepts images, video, and text and generates text. It pairs a 120B-parameter hybrid Mixture-of-Experts language backbone (12B active parameters per token) with a C-RADIO vision encoder, supports up to 1 million token context, and is available in Amazon SageMaker JumpStart.CURRENT SNAPSHOT3/5 DIMENSIONS WITH DATA
The dimensions that change the decision.
Maximum context lengthSupports context lengths of up to 1 million tokens.
Modality and I/OSuper VL is a vision-language model (VLM) that accepts images, video, and text as input and generates text.
Not established from the available sources.
Availability in Amazon SageMaker JumpStartNVIDIA Nemotron 3.5 Super VL is available in Amazon SageMaker JumpStart and can be deployed from SageMaker Studio or via the SageMaker Python SDK.
VERIFIABLE FACTS
Every value stays attached to a source and date.
AVAILABILITY · Availability in Amazon SageMaker JumpStartDEVELOPER CLAIM
NVIDIA Nemotron 3.5 Super VL is available in Amazon SageMaker JumpStart and can be deployed from SageMaker Studio or via the SageMaker Python SDK.
MODALITIES · Modality and I/ODEVELOPER CLAIM
Super VL is a vision-language model (VLM) that accepts images, video, and text as input and generates text.
CAPABILITIES · Intended applicationsDEVELOPER CLAIM
Built for document intelligence, GUI understanding for computer-use agents, video question answering, and multimodal agentic applications.
CAPABILITIES · Model size and active parametersDEVELOPER CLAIM
Nemotron 3.5 Super VL is a 120B-parameter model with 12B active parameters per token.
CAPABILITIES · Language backbone architectureDEVELOPER CLAIM
Language backbone has 88 layers that interleave Mamba-2 and MoE layers (512 routed experts, 22 active per token) with 8 attention layers and a shared Multi-Token Prediction (MTP) head; backbone initialized from Nemotron 3 Super Base and distilled from Nemotron 3 Ultra Base.
CAPABILITIES · Vision encoder and precisionDEVELOPER CLAIM
Includes a C-RADIO vision encoder that encodes images and sampled video frames; a vision-language connector maps encodings into the language backbone embedding space. The encoder and connector remain in BF16 in both provided variants.
CONTEXT WINDOW · Maximum context lengthDEVELOPER CLAIM
Supports context lengths of up to 1 million tokens.
CAPABILITIES · Multi-Token Prediction and speculative decodingDEVELOPER CLAIM
Includes a built-in Multi-Token Prediction (MTP) head that supports speculative decoding; reasoning can be configured per request.
WHAT CHANGED