NVIDIA Releases Nemotron 3.5 Lightning and NeMo Switchyard: Enhancing Efficiency and Control in Agentic AI
Summary:NVIDIA has launched Nemotron 3.5 Lightning, a highly efficient 30-billion-parameter MoE model designed for long-running agentic AI workloads, alongside NeMo Switchyard, an open-source smart routing library that intelligently directs requests to the most suitable model for each task. Nemotron 3.5 Lightning offers up to 4x faster output speeds compared to similar models and supports flexible deployment across local AI systems, data centers, and cloud environments, providing enterprises with a more
NVIDIA Releases Nemotron 3.5 Lightning and NeMo Switchyard
As AI transitions from chatbots to autonomous agents, open models are becoming increasingly crucial to meet market demands for full control over AI execution, deployment, and evolution. NVIDIA has expanded its Nemotron 3 model family with the launch of Nemotron 3.5 Lightning, a highly efficient 30-billion-parameter MoE model designed for long-running agentic AI workloads. Alongside this, NVIDIA released NeMo Switchyard, an open-source smart routing library that intelligently directs requests to the most suitable model for each task, enhancing the overall efficiency and control of AI applications.
Key Features of Nemotron 3.5 Lightning
- Efficiency: Nemotron 3.5 Lightning offers up to 4x faster output speeds compared to similar models, significantly improving the execution of agentic tasks.
- Customizability: The model can be post-trained using NVIDIA NeMo on an organization's own domain data, tools, and workflows to improve accuracy for specialized tasks.
- Flexibility: Nemotron 3.5 Lightning can run on local AI systems such as NVIDIA RTX PCs, DGX Spark, and Jetson, as well as scale across data centers and cloud environments, providing flexible deployment options for enterprises.
NeMo Switchyard's Capabilities
NeMo Switchyard is an open-source smart routing library that automatically routes prompts to the most capable and efficient model for each step of an agent workflow based on specific needs. Developers can tune or modify the router with different routing algorithms to match their priorities, such as quality, latency, and cost requirements, optimizing the performance of AI applications.
Industry Applications and Partnerships
Nemotron 3.5 Lightning and NeMo Switchyard have been applied across various industries:
- Cybersecurity: CrowdStrike leverages Nemotron 3.5 Lightning to enhance the accuracy of cybersecurity analytics.
- Legal Services: Harvey collaborates with Trajectory to optimize legal service intelligence with Nemotron 3.5 Lightning.
- Code Review: CodeRabbit and Baseten utilize Nemotron 3.5 Lightning to improve the efficiency and accuracy of code review.
- Scientific Research: Lila Sciences employs Nemotron 3.5 Lightning to boost reasoning capabilities in physical and life sciences.
- Software Development and Healthcare: Fastino Labs customized Nemotron 3.5 Lightning, achieving leading accuracy in software development and healthcare workloads.
Developer Support and Data Release
NVIDIA has also released Nemotron-RL-Agentic-Terminal-Pivot, an agentic reinforcement learning dataset used to post-train Nemotron 3.5 Lightning for coding agent capabilities. Additionally, the source code and training data for Nemotron 3.5 Lightning and NeMo Switchyard are publicly available within licensing constraints to ensure traceability, auditability, and support for training other models.
Conclusion
The release of Nemotron 3.5 Lightning and NeMo Switchyard marks another significant milestone for NVIDIA in the field of agentic AI. By providing efficient, customizable, and flexible AI solutions, NVIDIA equips enterprises and developers with powerful tools to tackle the complexities of modern AI applications.
— END —Source: NVIDIA AI News (2026-08-11)
Tags: #NVIDIA #Nemotron #Agentic AI #MoE Architecture #Open Source Model
Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.
Community Comments