AI BriefingGTC Unveils NVIDIA RTX PCs and DGX Spark for Local Execution of Latest Open Models and AI Agents
Key point
At GTC, NVIDIA expands the AI agent ecosystem by unveiling RTX PCs and DGX Spark that run the latest open models locally.
Details
The paradigm of personal computing is evolving beyond PCs and smartphones into an 'agentic computer' capable of privately running personal agents. To this end, NVIDIA introduced the DGX Spark desktop AI supercomputer and dedicated NVIDIA RTX PCs, signaling the arrival of the era of local AI agents.
A wave of next-generation open models designed to boost the intelligence of local agents was also unveiled.
- Nemotron 3 Super: A 120B parameter model optimized to run complex agentic systems using DGX Spark's 128GB unified memory.
- Mistral Small 4: A 119B parameter model optimized for general chat, coding, and agentic tasks.
- Nemotron 3 Nano 4B: A compact model that runs even in resource-constrained environments, suitable for building conversational personas in games or in-app experiences.
- Qwen 3.5: Through optimization of Alibaba's model, it offers vision support and a large context window.
A software stack to optimize the agent experience was also added. NVIDIA NemoClaw is an open-source stack that strengthens security and optimizes local model support, while Unsloth Studio makes fine-tuning models for agentic workflows even easier.
Additionally, Lightricks' audio-video model LTX 2.3 now supports NVFP4 and FP8 distilled models, improving performance by 2.1x.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Stanford and NVIDIA Release 'Contrastive Language Model' CLM with 9x Faster Inference