AMD first launched the Ryzen AI 400 series APU for mobile platforms at CES in early 2026, and later expanded the series to desktop platforms in March, both based on the latest Zen 5 architecture. Today, AMD has further enriched this product line and officially released the Ryzen AI Max+ 400 PRO series for the high-end and professional markets, succeeding the previous Max PRO 300 series.
The flagship model released this time is Ryzen AI Max+ PRO 495, which will be the first to be installed on the next-generation Ryzen AI Halo platform, positioning it as the core engine of local AI and high-performance computing. This processor adopts a 16-core 32-thread design with a base frequency of 3.1GHz and a maximum acceleration frequency of 5GHz. It integrates an 80MB cache and runs in a configurable TDP range from 45W to 120W to balance performance and energy consumption needs.
As a consistent feature of the Max series APU, the new generation 400 series still reserves a large amount of space in the chip area for integrated graphics. Ryzen AI Max+ PRO 495 has a built-in 40 computing unit (CU) GPU based on the RDNA 3.5 architecture and an integrated NPU based on the XDNA 2 architecture. The AI computing power can reach 55 TOPS, providing hardware support for local large-scale language model reasoning and various generative AI workloads.
In addition to the flagship model, AMD also simultaneously launched Ryzen AI Max+ PRO 490 and Max+ PRO 485, both of which have differentiated configurations in terms of CPU cores and GPU units. The 490 and 485 models adopt 12-core and 8-core designs respectively. The number of integrated GPU computing units is reduced from the flagship’s 40 to 32. The AI computing power of the built-in NPU is 50 TOPS, but the overall performance still meets the requirements of new generation AI personal computer platforms such as Copilot+ PC.
Compared to the previous generation Ryzen AI Max PRO 300 series, the new 400 series offers a significant jump in large model processing power and memory subsystems. In terms of large-scale language model inference, the 400 series can support a processing scale of more than 300 billion parameters, becoming the world's first x86 CPU product that can reach this level. In order to cope with the higher model scale and video memory requirements, the upper limit of the new series of memory has been increased from 128GB in the 300 series to 192GB, of which up to 160GB can be allocated as "class video memory", while the previous generation only supports up to 96GB.
AMD said that the Ryzen AI Max+ 400 PRO series will be officially launched in the third quarter of 2026, targeting high-end AI PCs, professional workstations, and desktop and mobile platforms that require powerful local AI inference capabilities, providing a stronger computing power foundation for Windows 11 and subsequent systems.