llama.cpp b10021 release includes DeepSeek V4 graph split reduction

The llama.cpp project has released version b10021, which incorporates a change to reduce graph splits for the DeepSeek V4 model.

This update provides pre-built binaries and frameworks across multiple platforms and hardware configurations. macOS builds are available for Apple Silicon (arm64) and Intel (x64), along with an iOS XCFramework. Linux support covers Ubuntu x64, arm64, and s390x architectures using CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL FP32, and SYCL FP16 backends. Android arm64 (CPU) binaries are also included.

Windows users can download builds for x64 and arm64 CPUs, as well as GPU support via OpenCL Adreno, CUDA 12.4, CUDA 13.3, Vulkan, OpenVINO, SYCL, and HIP. Additionally, openEuler builds are provided for x86 and aarch64 architectures using ACL Graph.