As of: 2026-10-15 23:34 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/m/nemotron-3-5-super-vl/

# NVIDIA Nemotron 3.5 Super VL (120B-A12B)

NVIDIA, Nemotron 3.5 family. Status: current. Type: multimodal model. Released: 2026-10-15. Training cutoff: not published.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Hugging Face |  |  | https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4 |
| OpenRouter | `nvidia/nemotron-3.5-super-vl` |  | https://openrouter.ai/nvidia/nemotron-3.5-super-vl |
| NVIDIA NIM / build.nvidia.com (free rate-limited dev API) |  |  | https://blogs.nvidia.com/blog/nemotron-3-5-super-vl/ |
| ModelScope |  |  | https://blogs.nvidia.com/blog/nemotron-3-5-super-vl/ |

## Pricing

- Input: $0.15 per 1M tokens
- Output: $0.50 per 1M tokens
- Unit as published: per 1M tokens (USD) on OpenRouter (also a rate-limited :free variant)
- Source: https://openrouter.ai/api/v1/models
- Last checked: 2026-10-15

## Specs

- Input: text, image, pdf
- Output: text
- Context window: 1,000,000 tokens
- Max output: not published
- Open weights: yes
- Licence: openmdw-1.1

## What it has not seen

- No published training cutoff. Counted from its release (2026-10-15): at least 0 AI events since (0 major, 0 historic), as of 2026-10-15.
- Everything since September 2026: https://postcutoff.com/since/2026-09/

## What stands out

- Open vision-language agent model with 1M context: 120B total / 12B active Mamba-2–Transformer hybrid MoE with a C-RADIO vision encoder and native multi-token prediction; reads PDFs, charts, tables, images and web pages; up to 1M tokens. Source: https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4
- Fast at long input: About 625 output tokens/s at 100K input on four B200s per Artificial Analysis (via VentureBeat); NVIDIA claims up to 5x the throughput of comparable open models. Source: https://venturebeat.com/technology/nvidias-new-open-nemotron-3-5-super-vl-with-post-training-beats-claude-opus-5-gpt-6-1-sol-on-specialized-task-benchmarks

## Notes

Trained on ~26T tokens, 3.8M post-training samples and RL in ~100 open environments. Ranks below GLM-5.3-Flash and Qwen3.8-Flash-Next on the Artificial Analysis index (~26). Production use via NVIDIA AI Enterprise ($4,500/GPU/year).

NVIDIA's open multimodal model for enterprise document and tool-using agents, released with training recipes and data for customization in NeMo.
</content>
</invoke>

Model page: https://postcutoff.com/m/nemotron-3-5-super-vl/