Configure Gigabyte AI TOP ATOM for business

Move AI TOP ATOM from installation to production use? WZ-IT provides the GB10 class with Open WebUI, a local model, hardening, integration, and support as the AI Cube Pro. Explore the AI Cube Pro

Gigabyte AI TOP ATOM is a compact AI computer based on NVIDIA's GB10 Grace Blackwell Superchip. Its 128 GB of unified memory and up to 4 TB of NVMe storage provide a foundation for local inference, model testing, and development. Gigabyte also provides AI TOP software tools. Company-wide use still requires defined user, data, and operating paths.

Understand the hardware

Gigabyte states up to 1 PFLOP FP4, models up to 200 billion parameters, and ConnectX-7 for connecting two systems. The 20-core Arm processor and Blackwell GPU use the same memory, so software and containers need ARM64 support.

Maximum parameter count does not equal a sensible business model. Quantisation, context, KV cache, and concurrent chats materially affect memory use and throughput.

Business setup

A controlled setup documents firmware and drivers; configures named administrators, SSH, DNS, TLS, and firewall rules; selects a runtime and model for the expected concurrency; deploys Open WebUI with users and model permissions; covers models, chat data, knowledge, and configuration in backups; and defines monitoring, updates, restore, and support access.

AI TOP software and Open WebUI

Gigabyte's tools primarily support development and local AI workflows. Employees usually benefit from one browser-based interface. Open WebUI provides chat, model selection, personal and shared knowledge spaces, and user management.

The platform can offer local and approved external models together. Technical and organisational rules must define which data may reach which endpoint.

When AI TOP ATOM is a good fit

AI TOP ATOM is suitable where Gigabyte is the preferred supplier or the accompanying AI TOP tools add value to development workflows. Open WebUI still provides the central interface for later use by business departments; developer tools and an employee platform serve different needs.

SSD configuration, target models, and concurrency should be defined before ordering. An internal knowledge assistant for one team has different priorities from a development group maintaining several model artefacts. Because the common GB10 foundation offers similar core performance, availability, storage, support, and integration often matter more than theoretical compute.

Complete entry and later scaling

The AI Cube Pro adds Open WebUI, model runtime, an agreed local model, hardening, functional testing, initial setup, and five hours of support. Automated RAG sources, interfaces, and MCP servers are added for the use case.

With growing use, another AI Cube can serve independent requests. Distributed inference is possible for larger individual models but is not automatically faster. The ConnectX-7 guide explains both paths.

Sources

• Gigabyte: AI TOP ATOM product page
• Gigabyte: AI TOP ATOM overview
• NVIDIA: GB10 and DGX Spark hardware