NVIDIA Unveils 64GB DGX Spark Desktop Computer Starting at $4,999
On October 2, NVIDIA announced an update to its DGX Spark lineup, introducing a 64GB memory version of its desktop AI computer at a starting price of $4,999, with availability set for October 23. OEM partners including Acer, Dell, ASUS, Gigabyte, MSI, and H3C will offer systems based on the platform.
The DGX Spark is powered by the GB10 Grace-Blackwell superchip with a unified memory architecture that bridges CPU and GPU memory. The new 64GB model provides about 56GB of usable memory for model weights and KV cache after reserving approximately 8GB for the system, supports NVFP4 quantization, and can run open-source models such as Gemma4 26B and Qwen3.8-27B. The device supports multi-node clustering via an integrated ConnectX-7 NIC and the NVIDIA Sync toolkit, enabling up to four units to be linked with automated network configuration and SSH setup.
In NVIDIA's tests, a two-node 64GB cluster running Qwen3.8-27B delivered up to 1.7x performance improvement over a single unit. The system includes the CUDA AI software stack, supports vLLM and llama.cpp, and is optimized for local agent workflows, with up to 1.9x faster local agent inference. Perplexity has also adapted its portable computer agent for DGX Spark. The 128GB FE version is now priced at $6,950. The platform is aimed at AI researchers, independent developers, and small AI startups.

