Nemotron 3.5 Super is the new mid-size flagship in Nvidia’s Nemotron 3.5 family — and the release that matters is the VL (vision-language) version, Nemotron 3.5 Super VL, which reads images and documents as well as text. Ollama even lists it simply as nemotron-3.5-super. My video below covers it. Here’s where it sits in the Nemotron line-up, what the names mean, and how to start using it.
Key takeaways
- Nemotron 3.5 Super is a 120B mixture-of-experts model with 12B active.
- The release is Super VL: text, images, documents and short video in one model.
- Other Nemotron 3.5 models include Lightning and a content-safety model.
- Run it with
ollama run nemotron-3.5-superor free via OpenRouter.
What Nemotron 3.5 Super is
Nvidia names its Nemotron models by size tier. “Super” sits between the small Nano-class models and the biggest Ultra models — big enough to reason well, small enough to run efficiently. Nemotron 3.5 Super VL is the version with vision built in, and Nvidia says it’s comparable with its largest reasoning model at roughly a fifth of the parameters.
| Nemotron 3.5 Super VL at a glance | |
|---|---|
| Made by | Nvidia |
| Released | 15 October 2026 |
| Size | 120B total parameters, 12B active per token (mixture of experts) |
| Architecture | Hybrid Mamba-2 + Transformer MoE with a C-RADIO vision encoder and multi-token prediction |
| Context | Up to 1M tokens |
| Inputs | Text, images, document pages and MP4 video up to 2 minutes (text output) |
| Reasoning | Off, medium or on; supports tool calling |
| Licence | OpenMDW-1.1 — open weights, commercial use allowed |
| Where to get it | Hugging Face, OpenRouter (incl. a free variant), Ollama, LM Studio, NVIDIA NIM, ModelScope |
The Nemotron 3.5 family
| Model | What it’s for |
|---|---|
| Nemotron 3.5 Super VL | The 120B vision-language model for agents |
| Nemotron 3.5 Lightning | A faster, cheaper model (listed on OpenRouter, including a free variant) |
| Nemotron 3.5 Content Safety | A safety classifier for moderating content |
| Nemotron Bolt 74B-A9B | A model that fits in 24GB of memory, released alongside Super VL |
This page focuses on Super; Lightning and the Bolt models are separate releases.
Want this working in your business, not just bookmarked? Using free open models in your AI agents is exactly what we build together inside the AI Profit Boardroom — 3,400+ members, four live calls a week, and a full one-hour DeepSeek Harness course in the classroom.
Prefer it mapped 1-on-1 first? Book a free strategy session and we’ll plan it for your exact situation.
Join the AI Profit Boardroom →Claim Your Free SEO Strategy Session →
Nemotron 3.5 Super vs the older Nemotron 3 Super
The previous generation, Nemotron 3 Super, was a text model. Nemotron 3.5 Super adds a vision encoder, so it can read screenshots, charts, document pages and short videos, and it improved on agent benchmarks — it even beats the bigger Nemotron 3 Ultra on Nvidia’s Hermes Bench (63.44% vs 56.33%). If you used Nemotron 3 Super before, this is the upgrade.
How to use Nemotron 3.5 Super
| Where | How | Cost |
|---|---|---|
| OpenCode | Pick Nemotron 3.5 Super VL in the model drop-down | Free (at time of writing) |
| OpenRouter | Model id nvidia/nemotron-3.5-super-vl or :free |
Free variant, or $0.15/M input and $0.50/M output tokens |
| Hermes Agent | Choose it via OpenRouter (or your local server) in the model picker | Free or pay-as-you-go |
| Ollama | ollama run nemotron-3.5-super (80–90GB) |
Free, needs big hardware |
| LM Studio | Download the GGUF (Q4_K_M, about 70GB) | Free, needs 70GB+ RAM |
| Hugging Face | BF16 and NVFP4 checkpoints for vLLM or SGLang | Free weights, server GPUs |
Who Nemotron 3.5 Super is for
- Agent builders who need a free model that can see.
- Hermes and OpenCode users looking for a fast backup model.
- Companies that want open weights they can host themselves.
- Researchers — Nvidia released training recipes and datasets too.
Let me be clear: this is not a frontier model. It’s not going to replace Claude Opus 5.5 for your hardest work. What it is: a free, open, fast model that can read screenshots, PDFs and charts and then act on them — which is exactly the job a lot of agent work starts with.
How to choose between Nemotron 3.5 models
If you want a model that can see and act — reading screenshots, PDFs and charts inside an agent — Nemotron 3.5 Super VL is the one. If you need something lighter and cheaper for high-volume text work, look at Nemotron 3.5 Lightning, which is also on OpenRouter with a free variant. If you’re moderating user content, the content-safety model is built for that job. And if you want to run something capable on a laptop with 24GB of memory, Nvidia’s Bolt 74B-A9B is designed for that.
What the 120B and 12B numbers mean
A mixture-of-experts model is made of many specialist sub-networks. Nemotron 3.5 Super has 120 billion parameters in total, but for each token it writes, only about 12 billion are used — the experts best suited to that piece of text. You get the knowledge of a large model with the speed and running cost of a much smaller one. That’s why it can be both capable and fast, and why it’s practical as a free hosted model.
How I’d test it this week
Open OpenRouter’s free variant or OpenCode and give it three jobs: a screenshot of a table to turn into a spreadsheet, a long PDF to summarise, and a small coding task. That shows you its vision, long-context and coding strengths in under an hour.
Then compare it with whatever model you use today. If it handles your visual jobs as well or better, put it in its own agent profile and route those jobs to it from now on.
Common questions about Nemotron 3.5 Super
Is Nemotron 3.5 Super the same as Super VL? The release Nvidia shipped is Super VL, the vision-language version; Ollama simply lists it as nemotron-3.5-super. Can it run on a normal laptop? No — even compressed builds are around 70–90GB. Is it good for coding? Solid, with 71.10 on SWE-bench Verified, though the best coding models score higher. Is it safe for business use? The OpenMDW-1.1 licence allows commercial use, but read it before building a product, and keep humans in the loop for consequential decisions.
The bottom line on Nemotron 3.5 Super
Nemotron 3.5 Super is Nvidia’s best mid-size open model yet, and the VL version is the one to use. Try it free on OpenRouter or in OpenCode, or run it locally if you have the hardware.
FAQ: nemotron 3.5 super
What is Nemotron 3.5 Super?
Nvidia’s 120B-parameter mixture-of-experts open model (12B active); the released version, Super VL, adds vision.
What does VL mean in Nemotron 3.5 Super VL?
Vision-language — it understands images and documents as well as text.
Is Nemotron 3.5 Super on Ollama?
Yes — ollama run nemotron-3.5-super (about 80–90GB).
Is Nemotron 3.5 Super better than Nemotron 3 Super?
Yes — it adds vision and scores higher on agent benchmarks.
Is Nemotron 3.5 Super free?
The weights are open, and there’s a free variant on OpenRouter.
Two ways I can help from here. Join the AI Profit Boardroom — it’s where your AI agent model setup gets built with 3,400+ members doing the same.
Or grab a free strategy session and bring your questions.
Join the AI Profit Boardroom →Claim Your Free SEO Strategy Session →
About Julian Goldie
I’m Julian Goldie — founder of Goldie Agency, best-selling author, and an AI educator with 428K+ YouTube subscribers. I’ve spent 10+ years in SEO and online business and I test every AI agent on my own work before I recommend it. Join the AI Profit Boardroom for the daily builds, or book a free strategy session to talk through yours.
Related reading
Last updated October 2026. This page is a living guide to nemotron 3.5 super — the facts here move fast and it is updated as they do.
