NVIDIA is advancing local AI capabilities with a series of updates, including new models, tools and features designed to enhance developer workflows and agent performance. The company highlighted advancements in its open-source ecosystem, emphasizing how these tools support the growing interest in local AI applications. NVIDIA's latest offerings aim to streamline the development and deployment of intelligent agents by improving performance, accessibility and customization options for developers and AI enthusiasts.
Among the key updates, NVIDIA introduced Nemotron 3.5 Lightning, a customizable open model that enables faster task completion and improved efficiency for always-on agents. The model is designed to run locally on various NVIDIA platforms, including RTX PCs, DGX Spark systems, and Jetson devices. It offers up to 4x faster token generation and 30% faster time to completion compared to other open models in its class. Developers can fine-tune the model to align with specific tasks, such as writing in a preferred style, learning a specialty or coding in a particular framework.
NVIDIA also announced new features for DGX Spark, including Google Chrome as a native ARM64 Linux build and a resource monitor for tracking CPU and GPU usage. These updates aim to improve the user experience and simplify workload management across multiple systems. The company emphasized its commitment to supporting open-source development and collaboration, working with partners like vLLM, Ollama and Unsloth to provide optimized deployment options for its models.
Source: nvidia
