Nvidia has released Nemotron 3.5 Lightning, a new open artificial intelligence model designed to make AI agents faster, more efficient and easier to customize.
The new model is aimed mainly at AI systems that need to perform a large number of tasks continuously. These can include using software tools, checking results, reviewing code, handling automated workflows and passing jobs between different AI agents.
Nemotron 3.5 Lightning uses a Mixture-of-Experts architecture and contains 30 billion parameters in total. However, only about 3 billion parameters are active during each operation. This design allows the model to provide the capabilities of a larger AI system while using less computing power for individual tasks.
Speed is one of the main areas Nvidia is focusing on with the Lightning model. The company says the model can deliver up to four times faster output in some comparable workloads. This could be useful for businesses running AI agents throughout the day, where even small delays can add significant cost and processing time.
The model also supports a context length of up to one million tokens. A large context window allows an AI system to work with longer documents, conversations, codebases and other information without quickly losing earlier context.
Another important part of the release is flexibility. Nvidia is providing model weights, training resources and customization tools so developers can adapt Nemotron 3.5 Lightning for their own requirements. Developers can fine-tune it for specialized tasks such as coding, business automation, data processing or industry-specific AI applications.
Nvidia is also offering optimized versions of the model for different types of hardware. This makes it possible to run Nemotron 3.5 Lightning on suitable local systems as well as larger data-center infrastructure. The approach could help companies keep certain AI workloads closer to their own systems instead of sending every request to a large external AI service.
Alongside Nemotron 3.5 Lightning, Nvidia has introduced NeMo Switchyard, an open model-routing system. Switchyard can decide which AI model should handle a particular task based on factors such as speed, cost and required capability.
This type of routing could become increasingly important as companies start using several AI models instead of relying on one large model for everything. Complex reasoning tasks may still need powerful models, while smaller and faster models such as Nemotron 3.5 Lightning can handle routine actions more efficiently.
The release also strengthens Nvidia’s position beyond the hardware used to train and run artificial intelligence. Nvidia is increasingly building software, models and development tools that companies can use to create complete AI systems around its computing platforms.
Nemotron 3.5 Lightning reflects a wider shift toward smaller, specialized AI models that can perform specific jobs quickly instead of using the largest available model for every request.
For developers and businesses building always-running AI agents, that balance between performance, speed and computing cost could make Nemotron 3.5 Lightning an important addition to the growing open AI model ecosystem.
