vs. a 4× RTX desktop
A 4-GPU desktop needs dedicated 20-amp circuits and pulls up to 1,800 W under load, battling thermal throttling and PCIe bottlenecks. A GX10 cluster pulls a fraction of that and runs from a standard office outlet.
Enterprise AI. Anywhere.
Put private AI hardware to work through VS Platform. Deploy AI tools, agents and supporting services from the solution catalog, then manage them across locations from one control plane. The Mini Desk Rack combines ASUS Ascent GX10 compute, networking and a private Ollama deployment in a compact system.
Standard wall outletNative Linux + CUDA12 months VS Platform included
2 × ASUS Ascent GX10Grace Blackwell · 256 GB unified memory
High-speed interconnect + 10G network10-inch rack · power and thermal management
Private AIOllama · agents · apps
VS PlatformDeployment · security · monitoring
Simplify AI at the edge.
- Accelerate AI projects. Arrives clustered, networked and ready to serve models.
- Enterprise security. TPM-backed node identity, role-based access and centrally managed policy.
- AI-ready for agents & LLMs. private Ollama deployment on day one
- Edge-to-cloud visibility. health, performance and security in one portal
- Automated operations. updates, provisioning and lifecycle handled centrally
- Built to scale. add nodes, add sites, same management plane
What's included
- (2) ASUS Ascent GX10 AI Supercomputers
- (1) High-speed interconnect cable (cluster)
- 8-port 10G Ubiquiti switch
- 10-inch mini server rack
- Patch panel & thermostat
- Power management
- 12 months VS Platform subscription
- Private Ollama local LLM deployment
- Deployment support
Complete hardware + software + services bundle. Subscription runs 12 months from time of purchase.
Pick the memory your models need.
Unified memory scales linearly with nodes. The sweet spot isn't raw single-user speed. It's running massive models and serving many concurrent agents efficiently, from a standard outlet.
256GB unified memory
123
- Max model size
- Up to ~400B parameters
- Example models
- Llama 3.1 405B · Qwen3-235B-A22B-FP4
- Power
- Standard wall outlet, no dedicated circuit
- Included
- 12 months VS Platform · private Ollama · deployment support
Same budget. No breakers tripped.
vs. a high-tier Mac Studio
Plenty of memory, but Apple Silicon. GX10 runs native Linux and NVIDIA's CUDA stack, so every open-source AI project, library and framework works without workarounds.
vs. cloud GPUs
Run models locally and choose which data stays within your network. VS Platform manages the deployment alongside the rest of your edge.
Private AI, wherever the work happens.
Enterprise AI labsRetailManufacturingWarehousingSecurity operationsRoboticsPrivate AI environmentsDistributed edge computing
Need a compact site appliance?
FlexPondi Pi5 connects distributed locations to VS Platform for centralized deployment, security, monitoring and management.
Explore FlexPondi Pi5