Long-running agents spend most of their time executing: calling tools, validating results and delegating work. Nemotron 3.5 Lightning is built for this high-volume execution, at a size that can run anywhere from an NVIDIA DGX Spark to the data center. See it running agentic