Open SourceHardware & Inference 🇷🇺 12.08.2026 06:01

Nvidia Launches Open-Weight Nemotron 3.5 Lightning Running on a Single GPU

NVIDIANVIDIA MetaMeta
Nvidia has introduced the lightweight Nemotron 3.5 Lightning open-weight AI model, which runs on a single GPU found in laptops or desktops. The company positions it as a tool for agentic applications and offers free downloads to enterprise customers, aiming to boost hardware sales.
In late July, Nvidia CEO Jensen Huang voiced support for open-source AI models, and now the company has unveiled the lightweight Nemotron 3.5 Lightning model, which requires only a single GPU in a laptop or desktop PC. Enterprise customers can freely download, use, and modify the model without approval. Nvidia benefits because users still need GPU hardware, which it sells. Huang believes open models give companies control, spur competition, lower prices, and enhance cybersecurity and national sovereignty. Nvidia joined Microsoft in a consortium promoting open-source models and software in cybersecurity. Meta founder Mark Zuckerberg also supported open-source models. Nemotron 3.5 Lightning is aimed at agentic applications and has been tested by CrowdStrike, CodeRabbit, and Harvey. It is available on Hugging Face and Nvidia's site, alongside the NeMo Switchyard utility for selecting optimal AI models. The lightweight variant was enhanced through distillation from larger Nemotron models.
Abbreviations
GPU = Graphics Processing Unit — графический процессор
Source: 3DNews — original
Our earlier posts on this topic ↓
Fresh news