Editor-Curated Robotics Industry Briefing - RobotToday

NVIDIA Expands Nemotron Family with 3.5 Lightning for Enhanced Agentic AI Performance

NVIDIA has introduced the Nemotron 3.5 Lightning, the most efficient model in its Nemotron series, designed for long-running agentic AI workloads. This 30-billion-parameter mixture-of-experts model enhances the performance of specialized tasks within multi-agent systems, reflecting NVIDIA's ongoing commitment to improving open models for better accuracy and speed. The significance of this release lies in its ability to provide enterprises with greater control over AI deployment and efficiency. With the introduction of NeMo Switchyard, an open-source library for smart routing, organizations can tailor their AI systems to direct requests to the most suitable models without needing to rewrite applications. This flexibility is crucial as modern agentic systems evolve into ensembles of specialized models. Looking ahead, NVIDIA's Nemotron 3.5 Lightning is already being customized by industry leaders for various applications, including cybersecurity and legal services. The model's ability to run on local systems and its open nature for post-training with domain-specific data further enhance its appeal. No further timeline was disclosed at the time of publication.