Developer Tools

llama.cpp b9351 expands hardware support with ROCm, Vulkan, and KleidiAI

New release adds ROCm 7.2, KleidiAI, and SYCL for faster local inference.

Deep Dive

llama.cpp's b9351 release provides builds for macOS (Apple Silicon, Apple Silicon with KleidiAI, Intel), iOS, Linux (Ubuntu x64/arm64/s390x CPU, Ubuntu x64/arm64 Vulkan, Ubuntu x64 ROCm 7.2, OpenVINO), Android (arm64 CPU), and Windows (x64/arm64 CPU, x64 CUDA 12/13, Vulkan, HIP). Intel SYCL builds are marked disabled.

Key Points
  • Adds official builds for ROCm 7.2, KleidiAI, Intel SYCL, and OpenVINO
  • Windows binaries include Vulkan, CUDA 12/13, and HIP support
  • Precompiled for iOS, Android arm64, and IBM s390x architectures

Why It Matters

Local LLM inference now runs faster on AMD GPUs, Apple Silicon, Intel GPUs, and more hardware.