The llama.cpp project has released version b10019, which introduces support for the fp16 kernel type in conv2d_dw operations via SYCL.

  • macOS and iOS builds are available for Apple Silicon (arm64) and Intel (x64), along with an iOS XCFramework.
  • Linux binaries cover Ubuntu x64 and arm64 CPU, s390x CPU, Vulkan, ROCm 7.2, OpenVINO, and SYCL FP32/FP16 variants.
  • Android arm64 CPU builds are provided.
  • Windows releases include CPU (x64/arm64), OpenCL Adreno, CUDA 12.4 and 13.3, Vulkan, OpenVINO, SYCL, and HIP binaries.
  • openEuler support for x86 and aarch64 architectures is listed as DISABLED in this release.
  • A standalone UI package is also included.