Three device lines for local AI
At GTC Berlin, NVIDIA showcased its local AI hardware together. RTX Spark, launched last week, runs personal agents on Windows PCs; DGX Spark targets developers on Linux; and DGX Station is built for enterprise teams running frontier workloads.
DGX Spark and clustering
64GB DGX Spark systems roll out from partners on Friday, October 23. They can run models including Nemotron 3.5 Bolt 74B, Qwen3.8 27B and Gemma 4 26B-A4B. With built-in ConnectX-7 networking and NVIDIA Sync Cluster Assistant, multiple DGX Sparks can be linked; manual configuration now supports clusters of up to eight units, delivering 1TB of unified memory.
Software and model support
Model Launcher, arriving in NVIDIA Sync later this month, downloads a suitable model and container and connects it to OpenCode. The October DGX Spark software update adds Wi-Fi Assistant and updated Skills. DGX Station improves manageability with seven new integrations, and its Windows version launches in December with support for manufacturing apps and agent harnesses.
Models and open workflows
NVIDIA is offering day-0 support for Nemotron, GLM, DeepSeek, Qwen, MiniMax and FLUX.3 models on its local hardware. The announcements also include new accelerations for llama.cpp and vLLM, a collection of checkpoints optimized for local hardware, and three open reference workflows for robotics. A new local coding model, Recursive Flint, and early-access NVIDIA Cloud Functions for DGX Station round out the list.
