Developer Launches XeMaya: Large Model Inference Engine for Intel Arc GPUs

On November 11, 2026, a post appeared on Reddit's LocalLLaMA community introducing a new tool from open-source developers called XeMaya. The author of the post explained that they had been trying to get some of the newer MoE models running on Intel Arc GPUs, primarily DeepSeek V4.1 Flash and GLM 5.3 Flash. Intel support for these models remains fairly limited, and although the author could get certain things working through patched llama.cpp builds, performance was far from impressive. To address this, developer fable 5.2 built XeMaya as a fork of XeStrata focused on running such MoE models natively on Intel Arc hardware through SYCL and Level Zero. According to the post, the underlying idea is to bring a particular kind of model support to Intel platforms, though the original description is cut off at that point and the remaining details could not be verified. At present, the verifiable information about XeMaya consists only of the above, and no other reports have followed up on its subsequent development.