llama.cpp b10211 updates Vulkan SDK 1.4.357.0, expands platform builds

ggml-org has released llama.cpp b10211, the latest iteration of the widely used open-source library for running large language models locally. The release focuses on infrastructure improvements, most notably updating the Vulkan SDK to version 1.4.357.0, which enhances cross-platform GPU acceleration for LLM inference. This ensures smoother execution on devices that rely on Vulkan, including many Windows, Linux, and Android systems.

The release also expands prebuilt binaries across a wide range of platforms. macOS users get Apple Silicon (with optional KleidiAI optimizations) and Intel x64 builds, while Linux supports x64, arm64, and s390x CPU variants, plus Vulkan, ROCm 7.2, OpenVINO, and SYCL FP32/FP16 GPU builds. Windows users receive CPU, CUDA 12.4 and CUDA 13.3 DLLs, Vulkan, OpenVINO, SYCL, and HIP versions. Android arm64 CPU builds and various openEuler ACL Graph variants round out the roster. These binaries allow developers and hobbyists to run cutting-edge models like Llama and Mistral locally without needing to compile from source, making local AI deployment more accessible across heterogeneous hardware.