Hacker News front pageAug 12, 2026, 03:35 UTC+8

NVIDIA ships Nemotron 3.5 Lightning and NeMo Switchyard for faster, smarter agent routing

Nvidia Nemotron 3.5 lightning and NeMo Switchyard

NVIDIA added a 30B-parameter MoE model, Nemotron 3.5 Lightning, to its Nemotron 3 family. It targets specialized tasks inside multi-agent systems, delivering 4x faster output and 30% faster agentic task completion than peers. It runs locally on RTX PCs, DGX workstations, and Jetson. The company also open-sourced NeMo Switchyard, a routing library that directs requests to the best model for each job without app rewrites. CrowdStrike, Harvey, and CodeRabbit are already using customized versions. The post does not disclose pricing or a release timeline.

Why it matters: Nvidia released a 30B MoE model positioned as a specialized worker in multi-agent systems, not a general-purpose model. The 4x output speed and 30% task acceleration claims are useful references, and Switchyard is open-sourced. But this is Nvidia's own blog with no third-party...

Read the original ↗Export Markdown

Take

The reason to click: Nvidia is drawing a sharper line inside agent architectures—big models plan, small models execute. Nemotron 3.5 Lightning is the executor: a 30B MoE model built for high-volume specialized tasks like code review and security alert monitoring. Nvidia claims 4x faster output and 30% faster task completion than peers, and it runs locally on RTX PCs, DGX workstations, and Jetson.

I'd discount the speed claims a bit until we see independent benchmarks. The post doesn't disclose pricing or a release date. CrowdStrike, Harvey, and CodeRabbit are already customizing it, but that's their internal work—not a sign you can download a ready-to-use model today.

NeMo Switchyard is the other piece worth watching: an open-source routing library that directs requests to the best model for each job without app rewrites. Think of it as a traffic dispatcher inside your agent toolchain. Useful if you're running model ensembles, but the post doesn't detail integrations with LangChain, CrewAI, or similar frameworks, so the real-world setup cost is unclear.

Overall, this is Nvidia staking a claim in agent infrastructure, not a product you can use tomorrow.

Source: Hacker News front page · blogs.nvidia.com
Published: Aug 12, 2026, 03:35 UTC+8 · Aug 12, 2026