# NVIDIA releases Nemotron 3.5 Super VL 120B with 1M-token context and speculative decoding

Digest AI · Generative AI & Models · published 2026-10-12T00:00:00Z · updated 2026-10-15T15:00:24Z

Canonical: https://digestai.news/story/nvidia-releases-nemotron-3-5-super-vl-120b-with-1m-token-context-and

## Summary

NVIDIA launched the **Nemotron 3.5 Super VL 120B-A12B-BF16** model, a vision-language model supporting text, images, and video inputs. It combines **Mamba-2, Mixture-of-Experts (MoE), and attention layers** with a **C-RADIO vision encoder**, enabling reasoning, coding, tool use, and long-context understanding. The model has **120B total parameters** and **12B active parameters**, with a **1M-token context window** and **Multi-Token Prediction (MTP) for speculative decoding** to speed up inference.

The model is optimized for **NVIDIA GPUs (Hopper/Blackwell)** and released under the **OpenMDW License 1.1**. NVIDIA provides deployment guides for **vLLM and SGLang**, with examples for text, image, video, and tool-calling tasks. Benchmarks cover **agentic capabilities, reasoning, multimodal tasks, and multilingual performance**, though full validation requires use-case-specific testing. The model is available on **Hugging Face** as of **October 15, 2026**, targeting developers for customization, domain adaptation, and inference.

## Key points

- Nemotron 3.5 Super VL 120B supports **1M-token context** with **text, image, and video inputs** via MTP speculative decoding
- Model combines **Mamba-2, MoE, and attention layers** with a **C-RADIO vision encoder** for multimodal reasoning
- Optimized for **NVIDIA Hopper/Blackwell GPUs**, released under **OpenMDW License 1.1** on Hugging Face

## Why it matters

This model pushes boundaries for **long-context, multimodal AI agents**, enabling developers to build systems handling **documents, videos, and code** without chunking. Its **speculative decoding** could reduce inference latency, while **open weights** lower barriers for research and enterprise adoption.

## Sources

1. [nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B model family released (base, NVFP4, BF16, and LiLiCorr speculative decoding)](https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16) (huggingface.co, 2026-10-12, primary source)
2. [Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models](https://blogs.nvidia.com/blog/nemotron-3-5-super-vl) (NVIDIA Blog, 2026-10-15, primary source)

Part of the developing story: [NVIDIA AI Model Rollout Saga Unveiling](https://digestai.news/thread/nvidia-releases-nemotron-3-5-super-vl-120b-with-1m-token-context-and) (2 stories)

## Cite

Digest AI, "NVIDIA releases Nemotron 3.5 Super VL 120B with 1M-token context and speculative decoding", 12 October 2026, https://digestai.news/story/nvidia-releases-nemotron-3-5-super-vl-120b-with-1m-token-context-and

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/nvidia-releases-nemotron-3-5-super-vl-120b-with-1m-token-context-and.json