xFasterTransformer provides a GPU FasterTransformer-equivalent solution for CPU-based LLM inference on Intel Xeon processors. It supports distributed inference across multiple sockets and nodes for running larger models, and offers both C++ and Python APIs spanning from high-level to low-level interfaces. The project supports popular model architectures including Qwen, DeepSeek-R1, ChatGLM, and LLaMA, and includes integration with vLLM for OpenAI-compatible serving.