NVIDIA Nemotron 3.5 Super VL 120B A12B BF16
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16
Open Source · chat · open-weights
Open
Alert me on changes
Context
1M
Max output
8.2K
Weights
Open
API $/1M
—
Modalities
text · image
Released
12 Oct 2026
License: other · nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16
AI summary
● machine-written
NVIDIA Nemotron 3.5 Super VL 120B A12B open-weights multimodal model released
NVIDIA Nemotron 3.5 Super VL 120B A12B is an open-weights vision-language model with 120B total parameters and 12B active parameters, supporting text and image inputs. The model features a 1M token context window and supports reasoning, tool use, and structured outputs. It is available from multiple providers with varying pricing and output limits.
What's new
- Mixture-of-experts architecture with 12B active parameters out of 120B total
- 1M token context window for extended reasoning and multi-document analysis
- Multi-token prediction (MTP) enables native speculative decoding
- Available across 5 providers with pricing ranging from free to $1.25 per 1M output tokens
- Supports explicit reasoning mode for complex problem-solving
Best for
Long-context document analysis and cross-document reasoningAgentic applications requiring tool use and structured outputsComplex reasoning tasks with temperature controlVision-language understanding with image and text inputs
Source: https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16