AMD has unveiled the Ryzen AI Max PRO 400 Series, the first x86 chip with 192GB of unified memory capable of running 300-billion-parameter language models locally. This development could reshape local AI inference, offering a viable alternative to cloud-based solutions.
TL;DR
- AMD's Ryzen AI Max PRO 400 Series introduces 192GB unified memory for local 300B-parameter LLM inference, per AMD.
- The series includes two models, the PRO 485 and Max+ PRO 495, with up to 160GB allocable as GPU VRAM.
- ASUS, HP, and Lenovo are expected to release systems with the PRO 400 Series in Q3 2026, according to AMD.
What happened
On May 20, 2026, AMD announced the Ryzen AI Max PRO 400 Series, a pair of mobile and workstation processors featuring up to 192GB of unified memory. This memory pool is shared between the CPU, GPU, and neural processing unit, enabling local inference of large language models.
The series includes two models: the Ryzen AI Max PRO 485 and the Ryzen AI Max+ PRO 495. Both models support up to 192GB of total memory, with up to 160GB allocable as GPU VRAM, according to AMD's announcement.
Why it matters
The PRO 400 Series could democratize access to large-scale AI models, allowing developers, researchers, and enterprises to run frontier-scale models locally without relying on cloud-based solutions. This could reduce costs, improve latency, and enhance data privacy, according to AMD.
The series' unified memory architecture sets it apart from traditional discrete-GPU setups, offering a more flexible and integrated solution for AI workloads. However, the actual performance of the PRO 400 Series will depend on independent benchmarks and reviews, as AMD's claims are based on marketing and launch claims rather than fully documented results.
Key facts
- AMD's Ryzen AI Max PRO 400 Series is the first x86 chip capable of running 300-billion-parameter language models locally, per AMD.
- The series features up to 192GB of unified memory, with up to 160GB allocable as GPU VRAM, according to AMD.
- The PRO 400 Series includes two models: the Ryzen AI Max PRO 485 and the Ryzen AI Max+ PRO 495.
- ASUS, HP, and Lenovo are expected to release systems with the PRO 400 Series in Q3 2026, according to AMD.
- The PRO 400 Series' unified memory architecture shares a single large memory pool between the CPU, GPU, and neural processing unit, according to AMD.
- AMD's claims about running 300B-parameter models are based on marketing and launch claims, not fully documented benchmark results.
- The PRO 400 Series is designed to target the gap between consumer graphics cards and data center GPUs, offering a solution for running large-scale AI models locally, per AMD.
Context
AMD's Ryzen AI Max PRO 400 Series is part of a broader trend in the AI hardware industry towards unified memory architectures. This approach allows for more flexible and integrated solutions for AI workloads, as seen in Nvidia's DGX Spark, Qualcomm's Snapdragon X2 Elite, and Apple's M5 Ultra Mac Studio.
The PRO 400 Series is AMD's latest step in its Ryzen AI Max lineup, which has focused on memory capacity as the key bottleneck for local AI inference. The series' unified memory architecture is an evolution of the approach used in AMD's earlier Ryzen AI Max chips, known by the code name Strix Halo.