Receive the GX10 as a complete AI platform? In the AI Cube, WZ-IT combines GB10 hardware with Open WebUI, local runtime, an agreed model, hardening, functional testing, and initial setup. Explore the AI Cube
ASUS Ascent GX10 is a compact AI computer based on NVIDIA's GB10 Grace Blackwell Superchip. Its 128 GB of unified memory enables local model classes that do not fit into typical individual GPUs. Business use still requires more than starting a model. Users, data, network, and operations must form one platform.
Understand the hardware
ASUS describes the GX10 with GB10, 128 GB of coherent unified memory, 10 GbE, and ConnectX-7 support. Its ARM64 platform differs from a classical x86 workstation with a discrete GPU. Containers, Python packages, and utilities need ARM64 support or a controlled build process.
A model limit from a specification is not a user or performance promise. Actual work depends on the exact model, quantisation, context, concurrency, and runtime.
Setup in eight steps
- record firmware, operating system, NVIDIA driver, and runtime;
- use named administrators, SSH keys, and least privilege;
- configure stable addressing, DNS, firewall, and separate user and admin paths;
- plan storage for models, chat, knowledge, and backup with headroom;
- install a runtime compatible with model, API, concurrency, and ARM64;
- configure Open WebUI with persistence, users, groups, model access, and TLS;
- test language, quality, context, latency, and concurrent requests;
- define updates, monitoring, backup, restore, support access, and documentation.
What users see afterwards
Staff do not interact with CUDA commands or model containers. They receive Open WebUI as a familiar chat interface, can select approved local models, work with files, and create personal or shared knowledge spaces according to their rights.
Automated sources, business software, custom RAG pipelines, APIs, and MCP servers can then be added to the same platform. They are not blanket parts of base setup.
AI Cube as the finished product
The AI Cube is WZ-IT's fully prepared product configuration in this hardware class. It includes hardware, Open WebUI, local model runtime, an agreed and tested model, hardening, functional testing, briefing and shipping to Germany, Austria and Switzerland. The AI Cube costs from €6,490 excl. VAT one-time for the 1 TB version, with a two-year manufacturer warranty. Patching, backup, monitoring and support are covered by AI Cube Care from €349.90 excl. VAT per month.
The system is usually ready within two weeks after configuration approval. Personal delivery and setup within Germany can be arranged. The first twelve months of AI Cube Care run with the device, after that Care is cancellable monthly. The device and root access belong to the customer; without Care, the AI Cube keeps running, then without updates, monitoring and support.
Expansion and clustering
A second AI Cube can serve independent requests or participate in distributed inference. These use different software. Two systems are not one computer with 256 GB of shared memory. See connecting AI Cubes with ConnectX-7.
Where rack mounting, redundant power, several discrete GPUs, or stricter availability are required, the path leads to AI Cube Custom or a GPU server.
