AMD Ryzen AI Halo Review: 128GB Unified Memory and Local AI Playbooks

Key points

• The AMD Ryzen AI Halo features 128GB of unified LPDDR5X memory, with up to 96GB assignable as VRAM for AI workloads.
• TechPowerUp highlights the 'Ryzen AI Developer Center' software, which includes 'Playbooks' for guided local AI tasks like image generation and coding.
• The system ships with pre-installed models, including GPT-OSS-120B and Qwen3-Coder-30B, allowing immediate local inference.
• TechRadar reports a price of $3,999 for a similar 128GB configuration with 40 compute units and dual 10G LAN.
• Camera Jabber notes the Thunderobot Station, a competitor using the same processor, is priced at CNY 26,999 (approx. £2,800) in China.
• The device supports remote access via SSH and includes tools like Lemonade Server for OpenAI-compatible local LLM routing.

Background

This review follows the launch of several high-memory AMD Ryzen AI Max+ platforms in 2026. Earlier this month, AMD released a 192GB variant of the Halo platform, and other OEMs like GMKtec and Lucebox have launched similar workstations with 192GB to 224GB of memory. Microsoft’s Project Zenith, announced in September, also targets this hardware class by standardizing a 'ready-to-code' Windows 11 environment for devices with at least 64GB of unified memory.

How outlets are covering it

TechPowerUp emphasizes the software ecosystem, praising the 'Playbooks' and pre-installed models for reducing the historical friction of running AI on AMD hardware. TechRadar focuses on the hardware specifications and the $3,999 price point, noting the inclusion of dual 10G LAN and 40 compute units. Camera Jabber provides a market comparison, highlighting the Thunderobot Station as a cheaper alternative in China with similar specs but a smaller chassis and an SD card reader. While TechPowerUp notes the unified memory split is a software convention, TechRadar and Camera Jabber both confirm the 128GB capacity and 96GB VRAM allocation as key selling points for local AI inference.

Why it matters

The Ryzen AI Halo represents a shift toward local-first AI computing, offering a viable alternative to cloud-based inference for developers and enterprises. By bundling software, models, and hardware, AMD aims to lower the barrier to entry for local LLMs and image generation. The availability of multiple OEM variants at different price points suggests a growing market for high-memory mini-PCs designed for privacy and low-latency AI tasks.

What to watch

Watch for international availability and pricing of the Thunderobot Station and other Ryzen AI Max+ platforms. Monitor the expansion of AMD's 'Playbooks' catalog and the adoption of Lemonade Server by third-party applications. Expect further competition from OEMs as the market for local AI workstations grows, with potential price reductions as production scales.