AMD’s Ryzen AI MAX+ PRO 495 Powers GMKtec’s EVO-X5 PRO To Run 300B+ Models Locally On 192 GB of Unified Memory

Recent coverage from outlets including Wccftech, HotHardware, igor'sLAB, Tom's Hardware, techpowerup.com, and tech-insider.org details a significant hardware trend emerging around localized artificial intelligence processing capabilities. Specifically, the AMD Ryzen AI MAX+ PRO 495 processor—also referred to in coverage as Gorgon Halo or Gorgon Point—is being paired with 192 GB of unified memory. This hardware combination is specifically engineered to enable users to run large language models exceeding 300 billion parameters locally on compact desktop systems. Publications like Tom's Hardware note that Minisforum launched its local AI solutions at IFA 2026, showcasing hardware such as the AI Agent NAS N5 and the AI Mini Workstation MS-S1 utilizing the AMD processor. Other manufacturers are also entering this space with specialized compact designs.

CHUWI introduced the UniBox AI495 Pro, which packs the same memory and processor configuration into a 2.9-liter box. Meanwhile, GMKtec revealed the EVO-X5 PRO powered by the same AMD components, and Acemagic displayed its F2A and F7A mini PCs featuring options for Intel Panther Lake alongside AMD Gorgon Point silicon. Pricing details have also begun to emerge in reports, with igor'sLAB noting that certain Gorgon-Halo mini PC configurations are reaching up to 8,000 euros. This hardware wave addresses a growing demand for local artificial intelligence deployment, shifting computational loads away from cloud infrastructure directly onto desktop and workstation hardware. By combining high-capacity unified memory with specialized processors, these compact devices bridge the gap between enterprise-grade AI workloads and desk-side form factors.

The integration of 192 GB of RAM into systems with footprints as small as 2.9 liters represents a major shift in how much local processing power can be housed within miniature desktop enclosures. Coverage emphasizes that these setups are directly marketed for running massive language models without relying on remote server connections. As manufacturers continue to exhibit these systems at major technology showcases, attention turns to market availability and performance benchmarks. Current coverage does not yet specify exact release dates or broader retail distribution timelines for all announced units beyond their initial exhibition and launch windows. Observers will be tracking how these localized AI workstations perform under sustained workloads and whether additional manufacturers adopt the AMD Ryzen AI Max+ PRO 495 and 192GB memory standard in forthcoming hardware iterations.