ASUS Ascent GX10 AI Workstation

ASUS Ascent GX10 AI Workstation
ASUS Ascent GX10 AI WorkstationASUS Ascent GX10 AI WorkstationASUS Ascent GX10 AI WorkstationASUS Ascent GX10 AI WorkstationASUS Ascent GX10 AI WorkstationASUS Ascent GX10 AI WorkstationASUS Ascent GX10 AI Workstation

ASUS Ascent GX10 AI Workstation

$2,399.60$5,999.00

The ASUS Ascent GX10 AI Workstation is built on the NVIDIA DGX Spark GB10 platform and powered by the NVIDIA GB10 Grace Blackwell Superchip, pairing a 20-core ARM processor with 128GB of LPDDR5x unified memory and a 1TB NVMe M.2 PCIe 4.0 SSD. This stackable metal system is designed for AI developers running secure, long-running agentic workflows, with compatibility across frameworks such as OpenClaw and NemoClaw for private on-device inference, sandboxed execution, and governed data access.

Four USB-C 3.2 Gen 2x2 ports are available, one serving as power input for the included 240W adapter while the other three support compatible peripherals at up to 20 Gb/s, alongside an HDMI 2.1 output for external displays. Wi-Fi 7 and a 10 GbE RJ45 port handle high-speed networking, and the NVIDIA ConnectX-7 SmartNIC lets you link up to two Ascent GX10 systems to work with models up to 405 billion parameters. NVIDIA DGX OS comes preinstalled.

Key Features

Brand ASUS
Model GX10-GG0010BN
Product Type AI Workstation
Operating System Ubuntu, NVIDIA DGX
Processor NVIDIA GB10
CPU 20-Core: Up to 3.25 GHz Grace / Cortex A725 (10 Cores), Up to 3.8 GHz Grace / Cortex X925 (10 Cores)
GPU NVIDIA Grace Blackwell
Installed RAM 128 GB
RAM Type LPDDR5x
RAM Configuration / Capacity Onboard Only/Not Upgradeable
Total Capacity 1 TB
Storage Drive Type 1x 1 TB / M.2 NVMe SSD
Storage Expansion 1x M.2 Slot (PCIe 4.0) [Occupied]
Optical Drive No
External Bays No
Inputs/Outputs 1x USB-C 3.2 Gen 2x2 / Supports Power Delivery (180 W), 3x USB-C 3.2 Gen 2x2 / Supports Video Alt Mode and Power Delivery, 1x Proprietary
Display I/O 1x HDMI 2.1 Output
Audio I/O No
Network I/O 1x RJ45 (10 GbE)
Wi-Fi Wi-Fi 7 (802.11be) with MU-MIMO Support (2 x 2)
Bluetooth Bluetooth 5.4
Keyboard No
Pointing Device No
Built-In Speakers No
Built-In Microphones No
Media/Memory Card Slot No
Security Kensington Lock Slot
Power Supply 240 W with USB-C
Certifications BSMI, C-Tick, CB, CCC, CE, FCC, UL, VCCI
Dimensions 5.9 x 5.9 x 2" / 150 x 150 x 51 mm
Weight 3.26 lb / 1.48 kg

Highlights

  • NVIDIA DGX Spark GB10 Platform
  • NVIDIA GB10 Grace Blackwell Superchip
  • 128GB of LPDDR5x Unified RAM
  • 1TB NVMe M.2 PCIe 4.0 SSD

Materials & care

The ASUS Ascent GX10 AI Workstation is housed in a stackable metal chassis. Keep the vents clear and use the included 240W USB-C power adapter for power.

Compatibility

  • Compatible with frameworks such as OpenClaw and NemoClaw for private on-device inference, sandboxed execution, and governed data access.
  • Supports up to 200 billion parameter models on a single system, and up to 405 billion parameters when two Ascent GX10 systems are linked via the NVIDIA ConnectX-7 SmartNIC.
  • Three USB-C 3.2 Gen 2x2 ports support compatible peripherals at up to 20 Gb/s, and one HDMI 2.1 output supports external displays.
  • Wi-Fi 7 (802.11be) and a 10 GbE RJ45 port connect to compatible high-speed networks.

Warranty

Limited 1-Year Manufacturer Warranty

What's included

  • ASUS Ascent GX10 AI Workstation
  • Power Adapter & Cable
  • Limited 1-Year Manufacturer Warranty

Downloads

Frequently asked questions

What operating system comes preinstalled on the ASUS Ascent GX10?

The workstation ships with NVIDIA DGX OS preinstalled.

Can two ASUS Ascent GX10 systems be linked together?

Yes. The NVIDIA ConnectX-7 SmartNIC adapter allows up to two Ascent GX10 systems to be linked, enabling work with large models of up to 405 billion parameters.

What is included with the ASUS Ascent GX10 AI Workstation?

The box contains the ASUS Ascent GX10 AI Workstation, a power adapter and cable, and a limited 1-year manufacturer warranty.

Which AI frameworks are supported?

The system is compatible with frameworks such as OpenClaw and NemoClaw, supporting private on-device inference, sandboxed execution, and governed data access.