NVIDIA has announced a 64GB configuration of DGX Spark that ships with the DGX OS and the full NVIDIA AI software stack, designed to run capable local agents on-device without cloud dependency.
The 64GB SKU, sourced through partner manufacturers, is positioned to support models up to 100 billion parameters and to enable developers and researchers to run workloads directly on their hardware.
A COMPLETE LOCAL AI PLATFORM READY FROM DAY ONE
The 64GB DGX Spark configuration integrates NVIDIA Grace Blackwell compute, unified memory, NVIDIA ConnectX-7 networking and a CUDA-accelerated AI software stack in a single system. TechStaged has also covered Meta Llama Enterprise AI Stack Planning Guide.
It ships ready for agent development, including NVIDIA Agent Toolkit, CUDA-X AI libraries, Nemotron open models and popular runtimes such as Ollama, vLLM and PyTorch with CUDA.
- DGX OS and full NVIDIA AI software stack pre-installed
- Supports on-device agents, inference, fine-tuning, data science and edge development
- Blender and other creator applications are among early supporters with upcoming installers
SCALING OUT WITH NVIDIA SYNC CLUSTER ASSISTANT
Two DGX Spark 64GB units can cluster together to pool memory and expand model support beyond a single device.
A built-in NVIDIA ConnectX-7 NIC and a direct QSFP link enable memory pooling to 128GB and potentially larger parameter ranges, with performance improvements demonstrated in tests.
- NVIDIA Sync Cluster Assistant configures the multi-node setup automatically
- Clustered units run the same software stack to simplify scaling
- NVIDIA Sync Model Launcher will enable running Qwen-3.8 27B across single or clustered devices
AVAILABILITY, PRICING AND ECOSYSTEM
NVIDIA says DGX Spark 64GB will be available starting Friday, Oct. 23, with a starting price of $4,999, through partner manufacturers Acer, ASUS, Dell, Gigabyte, HP and MSI.
The platform ships with a path to local model deployment and supports several inference frameworks for quick boot-up and experimentation.
- Starting price: $4,999
- Launch date: Friday, Oct. 23
- Partners include Acer, ASUS, Dell, Gigabyte, HP and MSI
- Initial ecosystem support includes Blender and other local AI tools
RELATED COVERAGE
- Meta Llama Enterprise AI Stack Planning Guide
- NVIDIA GPUs Accelerate OpenAI’s GPT-6 Astra Ultrafast
- Cloudflare’s 2026 Founders’ Letter: AI-driven growth, automated traffic and a reimagined Internet
- Google expands AI & Economy program with Nobel laureates and senior economists
- AI & Automation articles
SOURCES
- NVIDIA Blog: NVIDIA DGX Spark 64GB Gives Developers More Ways to Build and Scale Local AI Published · Primary source
