NVIDIA GTC 2026 showcases a range of local AI announcements centered on running open models and AI agents privately on RTX PCs and DGX Spark hardware. Key highlights include new open models — Nemotron 3 Nano 4B and Nemotron 3 Super 120B — plus optimizations for Qwen 3.5 and Mistral Small 4. NVIDIA introduced NemoClaw, an open-source stack that brings local Nemotron models and the OpenShell runtime to OpenClaw, improving privacy and eliminating token costs. Unsloth Studio launches as a web-based UI for simplified model fine-tuning with support for 500+ models, QLoRA, LoRA, and full fine-tuning. Additional announcements cover RTX-optimized generative AI models (LTX 2.3 with NVFP4/FP8, FLUX.2 Klein 9B FP8), DLSS 5, and updates to NVIDIA AI for Media SDKs.
