This item is not available to buy online right now. Request a quote and our sales team will check availability and pricing for you.
New ASUS Ascent GX10 mini desktop AI supercomputer built on the NVIDIA GB10 Grace Blackwell Superchip. Up to 1 PFLOP AI performance, 128GB unified LPDDR5x memory and 1TB NVMe.
Estimate shipping cost
→
Weight
Price
Service
Reference rates for one parcel. Contact us for more details and an accurate rate.
Out of Stock
BrandASUS
SeriesAscent GX10
ModelGX10-GG0010BN
Device TypeMini Desktop AI Supercomputer (NVIDIA DGX Spark)
SuperchipNVIDIA GB10 Grace Blackwell Superchip
CPU20-core Arm v9.2-A (10x Cortex-X925 + 10x Cortex-A725)
Request a quote or add to cart — pricing available on enquiry for refurbished stock.
For international customers, please contact us
for personalised support and assistance.
ASUS Ascent GX10 Mini Desktop AI Supercomputer – NVIDIA GB10 Grace Blackwell 128GB/1TB – New | GX10-GG0010BN NEW
New ASUS Ascent GX10 mini desktop AI supercomputer built on the NVIDIA GB10 Grace Blackwell Superchip. Up to 1 PFLOP AI performance, 128GB unified LPDDR5x memory and 1TB NVMe.
The ASUS Ascent GX10 is a mini desktop AI supercomputer built on the NVIDIA DGX Spark platform and powered by the GB10 Grace Blackwell Superchip, uniting a 20-core Arm CPU with an integrated Blackwell GPU for up to 1 PFLOP of FP4 AI performance. With 128GB of coherent LPDDR5x unified memory and a 1TB NVMe PCIe Gen4 SSD, it lets developers, researchers and data scientists prototype, fine-tune and run inference on models up to 200 billion parameters right at the desk.
Key Features
Platform: NVIDIA DGX Spark
Superchip: NVIDIA GB10 Grace Blackwell
CPU: 20-core Arm v9.2-A (10x Cortex-X925 + 10x Cortex-A725)
Common questions about the ASUS Ascent GX10 Mini Desktop AI Supercomputer – NVIDIA GB10 Grace Blackwell 128GB/1TB – New | GX10-GG0010BN.
It is built on the NVIDIA GB10 Grace Blackwell Superchip, pairing a 20-core Arm v9.2-A CPU (10x Cortex-X925 + 10x Cortex-A725) with an integrated NVIDIA Blackwell GPU featuring fifth-generation Tensor Cores.
Up to 1 petaFLOP (1,000 TFLOPS) of FP4 AI performance, enough to run AI models up to 200 billion parameters on a single unit.
Yes. Using the onboard NVIDIA ConnectX-7 networking, two GX10 units can be stacked to scale toward 405-billion-parameter workloads.
128GB of unified, coherent LPDDR5x system memory and a 1TB M.2 NVMe PCIe Gen4 SSD.
It ships with NVIDIA DGX OS (an Ubuntu Linux based stack) and is agentic-AI ready out of the box.