The llama.cpp project has released version b9924, which includes a refactor of fused operations. This update also addresses platform-specific configurations, notably disabling the KleidiAI integration for macOS Apple Silicon builds.

  • Refactored fused operations within the core codebase.
  • Disabled macOS/iOS KleidiAI support due to issues tracked in pull request #23780.
  • Provided binaries for macOS (Apple Silicon and Intel), iOS, Linux (CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL), Android, Windows (CPU, CUDA 12/13, Vulkan, OpenCL, OpenVINO, SYCL, HIP), and openEuler.

The release ensures compatibility across a wide range of hardware architectures and backend implementations while removing problematic features.