Odaily Planet Daily News: AMD has announced "Day 0" support for Alibaba's Tongyi Qianwen (Qwen) latest-generation model, Qwen3.8 27B, enabling developers to run this large open-source AI model locally on AMD hardware on the day of its release.
AMD states that Qwen3.8 27B is a 27-billion-parameter dense model optimized for local AI development, continuing the Qwen series’ focus on code generation, practical task execution, scientific research, and long-context AI applications. This model can run on AI PCs and workstations equipped with AMD processors, or on a single AMD 32GB GPU, and is supported on AMD hardware platforms with over 24GB of variable graphics memory (VGM) or dedicated memory.
Initial testing by AMD shows that Qwen3.8 27B delivers high local inference performance on AMD platforms: up to 24.5 tokens/second on the AMD Ryzen AI Max+ 395 processor, and up to 51.8 tokens/second on a single Radeon AI PRO R9700 GPU. Tests were conducted on Windows using the llama.cpp Vulkan backend with Multi-Token Prediction (MTP) optimization enabled. AMD notes that further software and model optimizations will continue to enhance real-world performance.
Alibaba today announced the official open-sourcing of the Qwen3.8 series models, which are freely available for download, deployment, and use by all developers, research institutions, and enterprises. The newly open-sourced Qwen3.8-27B is a native multimodal dense model with only 27 billion parameters, surpassing the overall performance of Qwen3.7-Plus.