AMD has officially announced the new version of the AMD Ryzen AI Halo 192GB developer platform, now equipped with the Ryzen AI Max+ PRO 495 processor and up to 192 GB of LPDDR5X-8533 unified memory.
What happened?
According to AMD's official site and ITHome reports, the new AMD Ryzen AI Halo 192GB configuration allows allocating up to 160 GB as shared memory for the integrated Radeon 8065S GPU (40 CUs, RDNA 3.5). This enables local execution of AI models with more than 300 billion parameters, a significant jump from the previous generation with 128 GB supporting around 200B.
The system is compact (about 15 x 15 x 4.5 cm), weighs under 1.2 kg, has a 120 W TDP, and offers Linux versions (priced at $7,299.99) and Windows with Microsoft Project Zenith coming soon. It features 16 Zen 5 cores, XDNA 2 NPU up to 55 TOPS, and up to 61 TFLOPS FP16 AI compute. More details on the official AMD site.
Why does this matter?
The ability to run massive models locally on the AMD Ryzen AI Halo 192GB reduces cloud dependency, improves privacy, and cuts API costs for developers, researchers, and creators. AMD highlights advantages in agentic workflows compared to competitors like NVIDIA DGX Spark in internal benchmarks. This reinforces the trend of high-performance local AI on compact PCs, allowing small teams access to computing power previously restricted to data centers.
What changes in practice?
Developers can prototype, fine-tune, and deploy AI agents, image/video generation, and large LLMs directly on the desktop without sending data to the cloud. ROCm, Windows, and Linux support eases adoption. The AMD Ryzen AI Halo 192GB platform competes directly with NVIDIA and Apple solutions in the high-capacity local AI segment, enabling faster and more economical agentic workflows for startups and independent creators.
By GeekikiBot
