Z3-AIB-02, Inspur YuanNao AI Agent Workstation, 2x RTX 5090, OpenClaw Local AI Appliance

The Inspur YuanNao Z3 is a local intelligent agent workstation and Agent-in-a-Box appliance for private LLM inference and multi-agent workloads. This high-performance listing adds two RTX 5090 24GB GPUs to the NP3460G7-based platform and is positioned for 120B-class model inference with up to 100 concurrent sessions.

Why Z3

Dual-GPU agent appliance

Two RTX 5090 24GB GPUs provide the accelerator capacity for higher-throughput local LLM and agent workloads.

120B-class local inference

The high-performance Z3 positioning supports 120B-class model inference with up to 100 concurrent sessions; actual throughput depends on model, quantization, and workload.

Managed private AI stack

ClawManager, OpenClaw, and Hermes provide a managed local software stack for deployment, monitoring, and agent operations.

Best Fit Applications

Multi-agent AI services

Run multiple local agents and private AI services for teams with higher concurrency requirements.

Large-model inference

Use the dual-GPU configuration for 120B-class local model inference and knowledge workflows.

SMB-scale private AI

Deploy a self-contained AI appliance for research groups, development teams, and midsize business environments.

Compare Related Models

	Z3-AIB-02	Z3-AIB-01
Configuration	High-performance / dual-GPU listing	Standard / no-GPU listing
GPU	2 × NVIDIA RTX 5090, 24GB each	No GPU in this listing; platform supports up to 2 GPUs
Model Positioning	120B-class model inference with up to 100 concurrent sessions	Local agent deployment; model size depends on final GPU and memory configuration
AI Agents	10+ independent sandboxed agents	10+ independent sandboxed agents
Software Bundle	ClawManager + OpenClaw + Hermes	ClawManager + OpenClaw + Hermes
Recommended Use	Higher-throughput local LLM inference and multi-agent workloads	Private agent deployment, research, and SMB workloads