The llama.cpp project released version b9916, which includes a fix for an incorrect A-matrix index in the `simd_gemm()` function's scalar tail-column path for full row blocks.
- macOS/iOS builds provided for Apple Silicon (arm64), Intel (x64), and iOS XCFramework; KleidiAI on Apple Silicon is disabled.
- Linux binaries available for Ubuntu x64 and arm64 (CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16).
- Android arm64 (CPU) builds included.
- Windows releases cover CPU, OpenCL Adreno, CUDA 12/13, Vulkan, OpenVINO, SYCL, and HIP for x64 and arm64 architectures.
- openEuler support is partially disabled; available for x86 and aarch64 with ACL Graph.
- A standalone UI build is also provided.
This release ensures correct matrix indexing in specific SIMD operations while providing updated binaries across major platforms and hardware accelerators.