NVIDIA releases multimodal Nemotron 3.5 Super VL for AWS EKS
In short: NVIDIA has introduced Nemotron 3.5 Super VL, a 120B-parameter multimodal model featuring a 1M token context window. The model utilizes a hybrid Mamba-2 and MoE architecture to balance reasoning capabilities with efficient inference, and is now available for deployment on Amazon EKS and SageMaker HyperPod.
This summary was generated automatically by AI from AWS Machine Learning's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1New 120B MoE model supporting text, image, document, and video inputs.
- 21M token context length support.
- 3Optimized deployment patterns for Amazon EKS using vLLM and NVIDIA Dynamo.
- 4Native support for speculative decoding via Multi-Token Prediction (MTP).
| Parameter | Before | Now |
|---|---|---|
| Parameters | Not specified | 120B total, 12B active per token |
| Context window | Not specified | 1M tokens |
| Architecture | Not specified | Hybrid Mamba-2, MoE, and Attention |
Why it matters
This model offers a specialized balance for agents requiring both visual intelligence and long-context reasoning. The ability to use Mamba-2 layers for memory-efficient long sequences is particularly relevant for maintaining performance in complex voice and data-heavy agent workflows.
What it means for AI agents and contact centers
When evaluating this model, monitor TTFT (Time To First Token) and ITL (Inter-Token Latency) using your specific agent workloads. Compare its reasoning and multi-modal handling against existing LLMs in your pipeline to see if the sparse 12B activation provides a cost-to-performance advantage in production.
🧪 Worth Testing
Testing is advised for high-complexity agent tasks where visual input analysis combined with large document context is required for decision-making.
Sources
- AWS Machine LearningOfficialPrimary sourceOriginal article →„Accelerate agentic AI with NVIDIA Nemotron models on Amazon EKS and SageMaker HyperPod EKS“29 Oct 2026, 18:11
- Published by source
- 29 Oct 2026, 18:11
- Found by our system
- 29 Oct 2026, 18:23
- Summary generated
- 29 Oct 2026, 18:23
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy