NVIDIA Nemotron 3.5 Lightning and NeMo Switchyard: AI faster, smarter and with full control

NVIDIA has officially launched Nemotron 3.5 Lightning and NeMo Switchyard, two new open source solutions for agency AI. The goal is to meet the growing demand for faster, smarter and more efficient AI agents, offering developers full control over where and how to run these models. While technology giants compete for space in the sector, NVIDIA invests in open infrastructure to attract teams that want speed without giving up control.

A.I.A.: What is Nemotron 3.5 Lightning?

It is the new generation of NVIDIA language models, optimized for agency tasks — that is, systems that plan, execute and use tools autonomously. Unlike a traditional chatbot, which only answers questions, an AI agent needs to reason about a goal, divide the problem into steps, trigger APIs and evaluate the results in sequence. For this, Lightning was designed to deliver low latency and high quality responses, maintaining the computational efficiency needed to run on different infrastructures, from RTX workstations to DGX servers. All this with open weights, allowing development teams to adapt the model to specific use cases without relying on proprietary APIs.

NeMo Switchyard: Total control for AI developers

If Lightning is the engine, NeMo Switchyard functions as the control cabin. The tool allows orchestrating AI agents with greater precision, defining how each model should be executed, where data should be processed and what criteria should guide decisions. This is essential for companies that need to meet privacy, security and cost requirements. With Switchyard, the developer can switch between run modes and optimize performance without rewriting the agent code. The proposal is flexible enough to give creative freedom, but with the necessary tracks so that the AI does not escape the scope defined by the organization.

AI in Brazil: impact on the market and opportunities

The demand for agency AI is not exclusive to the United States. In Brazil, more and more companies seek to automate complex flows with agents that interact with internal systems, APIs and knowledge bases. Having access to open source models with robust orchestration tools reduces dependence on closed solutions and enables local deployments, including in sectors with strong regulation, such as health and finance. In addition, integration with NVIDIA’s infrastructure – which already makes high bills with data centers for AI – places the company as the backbone of this new wave, which also became an investment asset, as we have already shown here.

Adopt AI agency: guide for companies with NVIDIA solutions

NVIDIA provided Nemotron 3.5 Lightning and NeMo Switchyard as open source projects, accessible by the company's official blog and repositories. This means that any engineering team can download the models, test the orchestration and integrate into the AI pipelines themselves. For those who already use the NVIDIA ecosystem in stations RTX or data centers DGX, adoption tends to be even more direct, with immediate gains in latency and efficiency. The complete documentation is in the [original publication of NVIDIA](https