GTC Spotlights NVIDIA RTX PCs and DGX Sparks Running Latest Open Models and AI Agents Locally

The landscape of consumer computing has long been defined by personal devices, from traditional PCs to smartphones and tablets. However, the advent of generative AI, particularly OpenClaw, has introduced a new category of devices known as agent computers. These machines, such as the NVIDIA DGX Spark desktop AI supercomputer and dedicated NVIDIA RTX PCs, are specifically designed to run personal agents — privately and without incurring significant costs.

The Global Technical Conference (GTC) currently underway is highlighting a range of agentic AI announcements, including new open models for local agents and optimizations for existing ones. NVIDIA is showcasing advancements that enable users to run state-of-the-art AI models on their personal devices, bringing cloud-level intelligence to the local level.

Among the new open models being introduced are NVIDIA Nemotron 3 Nano 4B and Nemotron 3 Super 120B. These models are designed to enhance the capabilities of local agents, providing them with the intelligence to perform complex tasks on personal computers. Additionally, NVIDIA is optimizing Qwen 3.5 and Mistral Small 4, ensuring that these models can run efficiently on NVIDIA devices.

To further enhance the user experience, NVIDIA has developed NemoClaw, an open-source stack for OpenClaw that optimizes performance on NVIDIA hardware. This stack not only increases security but also supports local models, making it easier for users to leverage the full potential of their agent computers.

In addition to these technical advancements, NVIDIA is also focusing on making it simpler for users to fine-tune their models. Unsloth Studio, a tool designed for easier fine-tuning, allows users to improve the accuracy of open models for agentic workflows. This democratizes access to powerful AI capabilities, enabling a broader audience to customize their agents to suit their specific needs.

At the GTC, attendees have the opportunity to participate in hands-on workshops, such as the "build-a-claw" event in the GTC Park. This interactive experience, running daily until March 19 from 8 a.m. to 5 p.m., allows participants to work with NVIDIA experts to customize and deploy a proactive, always-on AI assistant using their preferred device. Whether they are technical enthusiasts or simply curious about the possibilities of agentic AI, attendees can name their agent, define its personality, and grant it access to the tools it needs. The result is a personal assistant that can be accessed from their preferred messaging app, offering a new level of convenience and personalization.

The next generation of local models, with larger context windows, is delivering intelligence that can run on personal computers. Combined with richer user context and powerful local tools, these advancements are unlocking new possibilities for AI-powered devices, particularly the NVIDIA DGX Spark. With its 128GB of unified memory, the DGX Spark supports models with more than 120 billion parameters, making it an ideal platform for running complex agentic AI systems.

NVIDIA's Nemotron 3 Super, released last week, is a 120-billion-parameter open model with 12 billion active parameters. It is specifically designed to power agents on the DGX Spark or NVIDIA RTX PCs, ensuring that users can take advantage of the full capabilities of these powerful devices.

In conclusion, the GTC is showcasing a transformative shift in the world of consumer computing, with agent computers like the NVIDIA DGX Spark and RTX PCs becoming the new standard for running personal AI agents. By bringing cloud-level intelligence to local devices and making it easier for users to customize and deploy their agents, NVIDIA is paving the way for a new era of personalized, AI-driven computing. This evolution not only empowers individuals with advanced AI capabilities but also democratizes access to cutting-edge technology, enabling a broader audience to experience the benefits of generative AI in their daily lives.