NVIDIA has expanded its Nemotron open model family with the highly efficient 30-billion-parameter Nemotron 3.5 Lightning
As artificial intelligence transitions from chatbots to autonomous agents, open models are meeting market demand for full control over deployment and evolution. NVIDIA is expanding its Nemotron 3 model family with Nemotron 3.5 Lightning, described as the highest-efficiency model in its class for long-running agentic AI workloads. This release follows Nemotron 3 Nano and demonstrates NVIDIA’s ongoing commitment to improving open models for better accuracy and speed. Built for specialized tasks within larger multi-agent systems, Nemotron 3.5 Lightning is a 30-billion-parameter mixture-of-experts model designed to power smarter and more efficient agentic applications.
Alongside the new model, NVIDIA is releasing NeMo Switchyard, an open source library for intelligent routing within popular agent tools. Enterprises can use it to build custom routers that direct each request to the most capable and suitable model across a mix of open, proprietary, and NVIDIA models, all without requiring developers to rewrite their applications. Together, these tools provide greater control over how AI is deployed, where it runs, and how efficiently it operates across PCs, workstations, data centers, and the cloud.
Modern agentic systems increasingly function as model ensembles where different models handle different tasks. While frontier reasoning models like Nemotron 3 Ultra or GPT-5.6 may plan and orchestrate workflows, smaller specialized models like Nemotron 3.5 Lightning execute targeted tasks such as code review, tool use, security alert monitoring, and answering billing questions. Developed with contributions from the Nemotron Coalition, which provided evaluation methodologies, inference software, and datasets, the model delivers up to 4x faster output speed and achieves 30% faster agentic task completion compared with other models in its class. Because it is open and customizable, organizations can easily post-train it using NVIDIA NeMo on their own domain data, tools, and workflows to improve accuracy for specialized tasks.
AI leaders across industries are already customizing Nemotron 3.5 Lightning for specific workloads. These include CrowdStrike for cybersecurity, Harvey with Trajectory for legal services, CodeRabbit with Baseten for code review, Lila Sciences for improving reasoning capabilities in physical and life sciences, and Fastino Labs, which has customized the model and reports leading accuracies for software development, finance, and healthcare workloads. The model also gives organizations control over privacy and deployment, running locally on systems like NVIDIA RTX PCs, NVIDIA DGX Spark, NVIDIA DGX Station, and NVIDIA Jetson, or scaling across edge AI devices, NVIDIA RTX PRO workstations, data centers, and cloud environments. It is particularly suited for high-volume, specialized tasks requiring fast responses.
Consistent with previous Nemotron launches, NVIDIA is publishing as much training data and techniques as licensing permits to enable traceability, auditing, and further model training. Alongside Lightning, NVIDIA is releasing Nemotron-RL-Agentic-Terminal-Pivot, an agentic reinforcement learning dataset used to post-train the model for coding agent capabilities. The company notes that relying on a single default model often leads to overspending or quality loss, while manual routing creates integration bottlenecks. NeMo Switchyard solves this by automatically routing prompts to the most capable and efficient model for each workflow step based on specific needs. Developers can tune the router with different algorithms to prioritize quality, latency, and cost requirements, allowing enterprises to build powerful AI agents with improved tokenomics. Internal benchmarks show that NeMo Switchyard maintains frontier-level accuracy while reducing task completion cost to nearly one-third of Opus 4.8 alone.
NVIDIA is collaborating with partners across the AI ecosystem to integrate intelligent model routing into the tools and platforms developers already use. Nemotron 3.5 Lightning is available on Hugging Face, ModelScope, OpenRouter, and build.nvidia.com as an NVIDIA NIM microservice, as well as through a broad ecosystem of NVIDIA Cloud Partners, post-training platforms, inference platforms, and cloud service providers. NeMo Switchyard is available on GitHub and will be coming to partner platforms soon.