Specifications
Models
Ryzen AI Max+ PRO 495, Ryzen AI Max PRO 490, Ryzen AI Max PRO 485
CPU Architecture
Zen 5
GPU Architecture
RDNA 3.5 integrated graphics
NPU
AMD XDNA 2 (Ryzen AI)
Max Unified Memory
192 GB LPDDR5X-8533
Max GPU Allocatable VRAM
Up to 160 GB
Model Support
300B+ parameters at 4-bit quantization
Platform Codename
Gorgon Halo
Commercial Launch
IFA 2026, Berlin — September 4, 2026
Software Stack
AMD ROCm, Windows and Linux
Target OEMs
ASUS, HP, Lenovo
Form Factor Class
Commercial AI PC / small-form-factor AI workstation
Ryzen AI Halo Platform
Dev platform with 128 GB and 192 GB configurations
Availability
Q3 2026 (OEM systems; HP mobile workstation Oct 2026)
Target Workloads
Local agentic AI, LLM inference, fine-tuning, data science
Memory Bandwidth
LPDDR5X unified pool (up to 8533 MT/s)
Overview
The AMD Ryzen AI Max PRO 400 Series is AMD's halo-tier processor family for local AI, codenamed 'Gorgon Halo'. Commercially launched at IFA 2026 in Berlin on September 4, 2026, it combines Zen 5 CPU cores with RDNA 3.5 integrated graphics and an XDNA 2 NPU, and is led by the Ryzen AI Max+ PRO 495.
The defining feature is memory: the platform supports up to 192 GB of LPDDR5X-8533 unified memory, of which as much as 160 GB can be allocated directly as GPU VRAM. AMD states that capacity is enough to run 300-billion-parameter large language models at 4-bit quantization entirely on-device.
AMD positions the family for commercial AI PCs, mobile workstations and small-form-factor desktops aimed at data science and AI development teams, and is expanding its Ryzen AI Halo developer platform to 192 GB with 160 GB VRAM. OEM partners including ASUS, HP and Lenovo are expected to ship systems through the second half of 2026.
Key Benefits
192 GB unified LPDDR5X memory with up to 160 GB allocatable as GPU VRAM; 300B-parameter LLM support at 4-bit quantization on a desktop footprint; a full ROCm software stack across Windows and Linux; and a Ryzen AI Halo developer platform for local agentic AI development.
Applications
Local agentic AI and LLM inference, on-device model fine-tuning, data science workstations, AI development and testing, small-form-factor enterprise AI servers, and privacy-sensitive offline AI.
Request a Quote — AMD RYZEN AI MAX PRO 400 SERIES — 192GB UNIFIED MEMORY LOCAL AI PLATFORM
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →