Advanced chat and reasoning
Analyze, write, summarize, and translate across 201 languages and dialects, with strong Arabic and English support.
Private and secure · air-gapped operation
Private AI System Built on NVIDIA GB10
Two connected ASUS Ascent GX10 computers with 256GB combined unified memory and a 400B-parameter multilingual AI model installed before pickup.
256GBcombined unified memory
400Bmultilingual AI model
8–15tokens per second
Single usernot for multi-user serving
Full priceKWD 4,950
Reserve the system
Expected ready-for-pickup time: 1 week
Installed model included
Qwen3.5-397B-A17B multimodal AI
An optimized quantized build runs across both computers for private local reasoning, coding, tool use, and understanding of text, images, and video in Arabic and English.
View the official model on Hugging Face ↗What can you accomplish locally?
From a private enterprise assistant to a local AI studio, while sensitive files stay on your own system.
Images and document intelligence
Read images, tables, charts, presentations, and scanned documents, then extract and reason over their contents.
Video understanding
Upload video for scene, event, and content analysis and ask questions about it; video input is officially supported.
Generate and customize images
Run FLUX or Stable Diffusion locally through ComfyUI, customize FLUX for your brand, and create images at 1024px and higher.
Coding agents and tools
Generate code, understand large repositories, and connect to approved business tools and files for multi-step workflows.
Long context, private data
A native context of up to 262K tokens supports large documents and codebases while data and models remain on local storage.
Expected response speed
8–15 output tokens per secondDesigned for one active user only, not simultaneous multi-user serving. Speed varies with context length and workload; final performance is measured during commissioning.
Simulation at 10 tokens per second
● 10 tok/sUser
Summarize the benefits of private AI for a financial institution.
AI
Fully installed and configured before pickup
- ✓Two ASUS Ascent GX10 computers powered by NVIDIA GB10
- ✓QSFP56 DAC cable and direct computer-to-computer connection
- ✓Network, MTU, and RoCE v2 configuration for the internal link
- ✓DGX OS, CUDA stack, and container environment installation
- ✓400B-parameter multilingual AI model installed across both computers
- ✓One local chat interface or workflow
- ✓Air-gapped operation tested without internet access
- ✓14 days of post-pickup technical support
Private and secure by design
- 01No internet required for normal inference
- 02Can be fully isolated from external networks
- 03Business data, prompts, and models remain on local storage
- 04Local accounts, firewall configuration, and remote-access lockdown
- 05Controlled offline media for future updates or model changes
Final device configuration
| Specification | Details |
|---|---|
| Model No. | ASUS Ascent GX10-GG0004BN |
| Part No. | 90MS0371-M00040 |
| Color | Stellar Grey |
| Operating system | NVIDIA DGX OS (Ubuntu Linux-based) |
| Processor (CPU) | NVIDIA GB10 Grace Blackwell Superchip, ARM v9.2-A CPU with 20-core architecture |
| Graphics (GPU) | Integrated NVIDIA Blackwell GPU (GB10) |
| AI performance | Up to 1 PetaFLOP FP4 per computer; up to 2 PetaFLOPS across the dual-system configuration |
| Memory | 128GB LPDDR5X unified system memory per computer; 256GB combined across both linked systems |
| Storage | 1TB M.2 NVMe PCIe 4.0 SSD per computer |
| Maximum storage support | Up to 4TB M.2 PCIe SSD through one slot per computer |
| Networking | NVIDIA ConnectX-7 SmartNIC |
| Ethernet | 10 Gigabit Ethernet (10GbE) |
| Wireless connectivity | Wi-Fi 7 (Gig+), Bluetooth 5.4, and 2×2 antenna configuration |
| USB ports | 3 × USB 3.2 Gen 2x2 Type-C (20Gbps, DisplayPort 2.1 Alt Mode) |
| USB-C PD port | 1 × USB 3.2 Gen 2x2 Type-C with 180W PD input (PD 3.1) |
| Video output | 1 × HDMI 2.1 and DisplayPort 2.1 through USB-C Alt Mode |
| Additional ports | 1 × NVIDIA ConnectX-7 SmartNIC and 1 × Kensington Lock |
| Power supply | 240W external power adapter |
| Dimensions (W × D × H) | 150 × 150 × 51 mm |
| Weight | 1.48 kg per computer |
| Supported AI models | Up to 200-billion-parameter models locally on one computer |
| Scalability | Dual-system linking through NVIDIA ConnectX-7 for AI models up to 405 billion parameters |
| Included accessories | AC adapter, power cord, user manual, and warranty card |
| Warranty | 1 year |
Payment and pickup
- ●The full KWD 4,950 is paid by KNET when ordering to confirm the purchase and begin system preparation.
- ●In-store pickup in Kuwait only; shipping and customer-site installation are not offered.
- ●The system is delivered fully installed, configured, and tested for local isolated operation.
- ●One-year manufacturer warranty on both computers.
- ●Your card-issuing bank may impose a daily or per-transaction KNET limit; confirm that at least KWD 4,950 is available before paying.
Frequently asked questions
Which model is installed?+
Qwen3.5-397B-A17B multimodal and multilingual AI is distributed across both computers and tested in Arabic and English.
Can it generate images and video?+
The system can generate images through an additional compatible generation model. It can also run a separate video-generation model, although this is slower and depends on the model, resolution, and duration. The included Qwen3.5 model understands images and video but does not generate media files itself.
How fast does it respond?+
Expected single-user generation is 8–15 output tokens per second. Results vary with context length and workload, and final performance is confirmed during commissioning.
Does the system require internet access?+
The AI does not require internet after the software and model are installed. The system can operate fully air-gapped; future updates or models can be introduced through controlled offline media.
Is it ready when collected?+
Yes. We install the operating system, local AI environment, and multilingual model, then validate both computers in Arabic and English before in-store pickup.
Do you offer shipping or on-site installation?+
No. This product is available for in-store pickup in Kuwait only. Shipping and customer-site installation are not offered.
Have questions about the AI model?
Contact us: 9550 9361
Complete your purchase
Enter your organization details and pay the full KWD 4,950 by KNET to confirm the order. We will configure and test the system before pickup from our store ↗.
Out of stock