halogen
Inference engine tuned to run Qwen3.8-27B at high precision on AMD Strix Halo GPUs.
Pluginfreeglobal
halogen is an inference engine with kernels written specifically for AMD Strix Halo GPUs and the Qwen3.8-27B model family, delivering higher-precision, faster inference than general-purpose runtimes achieve on that one piece of hardware by dropping portability and fallback layers. It targets developers running Qwen3.8 models locally on Strix Halo who want the fastest inference that hardware can produce.
Categories
Developer ToolsAI Infrastructure
Something wrong with this listing — dead link, not a real product, wrong info?

