Skip to content
NerdLevelTech

NVIDIA

Nemotron 3.5 Super VL

NVIDIA’s open vision-language model with a 1M window

Open weightsNewReleased: October 15, 2026Self-hostingHigh volumeVerified: October 24, 2026

What it is

NVIDIA’s open vision-language model: a 1,048,576-token context, text and image input, and $0.15/$0.50 per 1M tokens on the hosted API. The public Hugging Face release is the NVFP4 build.

NLT take

The open multimodal option for agent workflows: a 1M window, image input and a real hosted price. It is the cheapest way here to give an agent eyes without closing the weights.

Before you switch

Ollama lists the Nemotron family, not this Super VL tag yet. The Hugging Face repo we link is the NVFP4 quantized build, not the full-precision weights.

Cost for a real task

The same task on every model: 50,000 input tokens and 10,000 output tokens, at the base tier.

Task cost

$0.013

50K input + 10K output tokens

Input, per 1M
$0.15
Output, per 1M
$0.50
Cached input, per 1M
$0.020

Pricing source: OpenRouter listing

Specs

Maker
NVIDIA
Licence
Open weights
Released
October 15, 2026
Context window
1,048,576 tokens
Parameters
120B total / 12B active
Memory (local build)
NVFP4 — Quantized Hugging Face build
Input / output
Text · Image → Text

Watch it in action

Hands-on videos from creators we trust. They open on YouTube; we do not host or re-upload them.

From our coverage

Articles about this model or its family. A “Background” tag means we have not covered this exact release yet.

Sources & model links

Every number on this card traces back to one of these links. The maker page proves the positioning, the pricing page proves the rates, the Hugging Face repo proves the weights exist, and the Ollama page proves a local tag.

Official

Open weights

Run locally