GMKtec has launched the EVO-X5 Pro at IFA 2026 in Berlin, positioning the desktop system as capable of running a 300-billion-parameter large language model (LLM) fully offline.
The company said the system is powered by AMD’s Ryzen AI Max+ PRO 495 processor and is scheduled to launch on September 28. GMKtec also pointed to a growing relationship with AMD, with Michael Nordquist, vice president of client business at AMD, appearing at the launch event.
GMKtec said the EVO-X5 Pro received two awards at IFA 2026: PCWELT’s “Best of IFA 2026” and the SmartZone “BEST OF IFA 2026 Innovation Award”.
In remarks included in the release, Nordquist said the platform combines the Ryzen AI Max system with up to 192GB of unified memory to support local LLMs and “next-generation AI agents” on the desktop.
According to GMKtec, the EVO-X5 Pro’s specifications include 16 Zen 5 CPU cores, 32 threads, boost speeds up to 5.2GHz, and an XDNA 2 NPU delivering up to 55 TOPS. The system supports up to 192GB of LPDDR5X-8533 unified memory, with the company stating memory bandwidth of 273GB/s and up to 160GB that can be allocated as graphics memory.
The company framed the offline LLM capability as reducing reliance on cloud infrastructure, lowering latency, improving data privacy, and reducing operating costs. It said the device is aimed at AI developers, enterprise deployments, research organisations, and professional content creators.
On security and manageability, GMKtec said the EVO-X5 Pro is built on the AMD DASH management framework and includes an RJ45 DASH network adapter and a dedicated dTPM 2.0 security chip. The company said this combination enables hardware-level security and out-of-band remote management independent of the operating system, and that the system has been validated for 24/7 continuous enterprise operation.
GMKtec said that during IFA 2026 it demonstrated an offline 300B-parameter LLM running on the device at its booth. The company also said it plans to deepen work with AMD across the ROCm software ecosystem, AI software optimisation, and what it described as the “Agentic PC” platform.
