TNG Technology Consulting released NVIDIA Nemotron 3.5 Super 120B VL on October 15, 2026, saying it extends the model's capabilities for agentic web development tasks. It is the company's first major update to the Nemotron family since the release of Nemotron 3.5 in 2025.
The model reported a 120B parameter count, measured on a multi-modal benchmark, with vision capabilities for visual tasks. That compares with the 65B parameter count of its predecessor, Nemotron 3.5.
Nemotron 3.5 Super 120B VL is built on a hybrid attention and Mamba architecture and targets web development tasks like generating, editing, and repairing web pages. Availability begins with early access for TNG Technology Consulting's clients and consultants.
"A state-of-the-art 120B model with a million token context," said Henrik Klagges, TNG Technology Consulting. "This model is highly efficient due to its hybrid attention and Mamba architecture."
The announcement follows TNG Technology Consulting's collaboration with NVIDIA on model training and benchmarking. TNG Technology Consulting framed the significance as a step toward more efficient and adaptable agentic systems.
NVIDIA did not say how the model will perform in other development frameworks like Angular or Vue, and the open question is whether the model can adapt to different agentic environments. TNG Technology Consulting plans to continue refining the model's web development capabilities through reinforcement learning.
Source: huggingface
