NVIDIA Announces DGX Spark 64GB: A 1-PetaFLOP Grace Blackwell Desktop for Local AI Agents, Fine-Tuning, and Inference
NVIDIA announced a 64GB DGX Spark desktop with up to 1 petaFLOP FP4 for local agents and fine-tuning.
NVIDIA announced a 64GB DGX Spark desktop, offered through Acer, ASUS, Dell, Gigabyte, HP, and MSI, built on the GB10 Grace Blackwell superchip. It provides up to 1 petaFLOP of sparse FP4 compute, 64GB of coherent LPDDR5x at 273 GB/s, ConnectX-7 200GbE networking, and DGX OS with PyTorch, Jupyter, and Ollama. NVIDIA says it fits today’s 30–35B-class open models, while two systems can cluster for 128GB of memory; the existing 128GB Spark remains available. Availability is listed for October 23, 2026.
- GB10 pairs a Blackwell GPU with a 20-core Grace Arm CPU.
- 64GB unified LPDDR5x is shared by CPU and GPU over NVLink-C2C.
- NVIDIA targets roughly 30–35B open models, including quantized Muse Glimmer.
- Two 64GB units can cluster for 128GB and up to 2 PFLOPS.
- Sales begin October 23, 2026, through NVIDIA and OEM partners.
From the original article
6 min read at marktechpost.comNVIDIA announced a new 64GB configuration of DGX Spark — from Acer, ASUS, Dell, Gigabyte, HP and MSI — its GB10-powered desktop AI system. It gives developers a way to start with one system for local models and agents, then cluster two 64GB units for 128GB of memory across the cluster and more compute when […]
By Jean-marc Mommessin · marktechpost.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.