Nemotron 3.5 Super VL
NVIDIA's open 120B multimodal reasoning model for agents, sized for a single GPU
Context window
1M
Input / 1M tokens
$0.10
Output / 1M tokens
$0.40
Provider
NVIDIA
Open weights under OpenMDW-1.1 on Hugging Face. Hosted price shown is OpenRouter (nvidia/nemotron-3.5-super-vl); a free variant is also listed. Also on build.nvidia.com. · Data verified 2026-10-17
Nemotron 3.5 Super VL (October 15, 2026) is an open-weight 120B-parameter model with 12B active per token, using a hybrid Mamba-2 and Transformer Mixture-of-Experts design with multi-token prediction. It reads text, images, documents and video, supports up to 1M tokens of context, and is built for specialized, always-on agents.
Capability index
Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.
How to access it
Hugging Face weights, build.nvidia.com, or OpenRouter.
Strengths
- ✓Very cheap hosted inference ($0.10/$0.40)
- ✓Fits on a single GPU for self-hosting
- ✓Multimodal input including video
Best for developers who...
Benchmarks
| Benchmark | Score | Notes |
|---|---|---|
| GPQA Diamond | 89.52 | NVIDIA-reported |
| SWE-bench Verified | 71.10 | NVIDIA-reported |
| Artificial Analysis Intelligence Index | 26 | Third-party |
Source: Hugging Face model card
Compare Nemotron 3.5 Super VL
Compare Nemotron 3.5 Super VL with any other model
Build a comparison →All model comparisons →