AMD has released its latest Gorgon Halo developer platform, featuring the Ryzen AI Max+ Pro 495 processor and 192GB of unified memory. Designed for local AI workloads, the system supports 300-billion-parameter models and is now shipping to developers and creators, following a series of preemptive benchmark releases ahead of the RTX Spark launch.
Ryzen AI Max+ Pro 495 and the 192GB Memory Ceiling
The new Gorgon Halo platform centers on the Ryzen AI Max+ Pro 495, an x86 processor that retains the core count and architecture of its Strix Halo predecessors while pushing memory capacity significantly higher. With 192GB of LPDDR5x unified memory, the chip allows for the local execution of large language models that were previously restricted to cloud-based environments. According to official AMD technical documentation, the system pairs 16 “Zen 5” CPU cores with 40 RDNA 3.5 GPU compute units and a 55 TOPS NPU to handle agentic workflows that require both orchestration and high-speed inference. The processor features a 100MHz increase over the previous generation.
This memory expansion provides a tangible increase in VRAM availability for professional users. While earlier Strix Halo iterations supported 128GB, the move to 192GB allows systems to assign up to 160GB of memory directly to the GPU, a 66% increase in available VRAM for compute-heavy tasks. The platform’s memory bandwidth reaches 273 GB/s, facilitated by 8.5 billion data transfers per second. A developer platform is currently listed at $7,299.
Hardware Deployments from Minisforum, iBUYPOWER, and OneXPlayer
System integrators have already begun integrating the Gorgon Halo silicon into compact form factors. The Minisforum MS-S1 MAX-P495 targets the professional workstation market with its 192GB configuration. Pricing for these units varies; while initial promotional discounts for some configurations have dipped near $7,000, standard pricing for high-tier models often exceeds $7,000.
Meanwhile, iBUYPOWER is targeting the mini-PC space with its Pro AMD AI Halo unit. For portable users, the OneXPlayer Super X Pro introduces a 2-in-1 design that supports an external liquid-cooling dock, capable of sustaining 120W of power for intensive AI model inference. The Super X Pro also includes a 2.8K AMOLED display at 2880 x 1800 resolution with a 120Hz refresh rate and support for HDR and variable refresh rates.
Lucebox has also entered the market with the Zero 495, a compact AI workstation that pairs the Ryzen AI Max+ Pro 495 with a separate AMD Radeon AI PRO R9700 GPU carrying 32GB of dedicated GDDR6 memory. The device utilizes static assignment at model load time to avoid the bottleneck of PCIe links between memory pools. The developer tier starts at $5,999 for the first 200 units, while the business tier reaches approximately $10,004 USD.
Competitive Positioning Against Nvidia’s RTX Spark
Throughout early October, AMD shared internal benchmarks comparing its new silicon against Intel’s Core Ultra X9 388H, as the RTX Spark hardware remained unavailable for third-party testing. Nvidia’s RTX Spark is expected to launch in two configurations: a 20-core CPU with 6,144 CUDA cores and an 18-core CPU with 5,120 CUDA cores, topping out at 128GB of memory.
Performance benchmarks for the Ryzen AI Max+ Pro 495 show up to 20 tokens per second when running the GLM 5.3 Flash model with 320 billion parameters. While AMD emphasizes its memory advantage, performance outcomes remain nuanced. As of October 13, 2026, independent reviews comparing production-ready RTX Spark units to the Gorgon Halo developer platform are still forthcoming.