Nemotron 3.5 Super VL with Native Speculative Decoding @fahdmirza
Nemotron 3.5 Super VL with Native Speculative Decoding  @fahdmirza
Uploaded October 2026 | Updated November 2026, 3 days ago
This video tests NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16, a vision-language model.

📬Weekly AI Newsletter: fahdmirza.substack.com

🔥 Get 50% Discount on any A6000 or A5000 GPU rental, use following link and coupon:

bit.ly/fahd-mirza
Coupon code: FahdMirza

🔥 Buy Me a Coffee to support the channel: ko-fi.com/fahdmirza

#nemotron #nemotronsupervl

PLEASE FOLLOW ME:
â–¶ LinkedIn: / fahdmirza
â–¶ YouTube: / @fahdmirza
â–¶ Blog: fahdmirza.com

0:00 The buggy LUMEN Rx app and the silent bug
2:52 What Nemotron 3.5 Super VL is
4:00 Bug fixed and verified
5:16 Test 2: Portuguese image to 82 language titles
6:59 Test 3: Circuit diagram scientific reasoning
8:15 Test 4: Dragon tattoo in pure HTML
9:35 Benchmarks and verdict

RESOURCES:

â–¶ huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16

All rights reserved © Fahd Mirza
Nemotron 3.5 Super VL with Native Speculative DecodingRun 35B Model on Phone Under 3GB Memory with Edge0Colibri Tutorial: Run Big Models Like DeepSeek, Qwen on One GPU LocallyObscura + Ollama  Local AI Web Scraping with a Rust Headless Browser5x Faster MiniMax H3 Locally with the Turbo LoRA and ComfyUINorth Micro Vision: A 2.4B Native-Resolution Vision Model: Run LocallyDots.TTS SOAR - State of the Art Speaker Similarity, Runs Fully LocalHumanizer 12B: Can Local AI Fix AI Slop? (Hands-On Test)Ripple-65B: Thin and Deep Dense Model, Tested LocallyRun Every AI Model Your Agent Needs From One Open-Source Server (SIE)Fine-Tune Qwen3.8 27B on Your Own Dataset Locally: Step-by-Step TutorialMesh LLM: Pool GPUs Across Machines Into One Local API
Fahd Mirza |

Nemotron 3.5 Super VL with Native Speculative Decoding

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER