RyzenClaw and RadeonClaw bring OpenClaw directly to your local AMD hardware. Optimized for Ryzen AI Max+ and Radeon PRO GPUs via WSL2, you can now smoothly run massive models like Qwen 3.5 122B entirely on-prem.

Instella, from AMD, is the high-performance 3B language models. ResearchRAIL license for model weights, MIT license for code. Trained on MI300X.