llama.cpp Releases Version 0.5.0 (Build b11146)
llama.cpp has officially bumped its version to 0.5.0 (build b11146) via pull request #29333, releasing updated pre-compiled binaries across multiple operating systems. As a foundational tool for running open-source large language models locally, milestone updates like version 0.5.0 ensure cross-platform compatibility across a wide range of consumer and enterprise hardware backends. The release binaries feature support for CUDA 12 and 13, ROCm, Vulkan, OpenVINO, and Snapdragon hardware acceleration across Linux, Windows, Android, macOS, and iOS. Notably, KleidiAI support for macOS Apple Silicon arm64 builds has been disabled for this build.
## BACKGROUND
llama.cpp is a popular open-source C/C++ engine that enables efficient local inference of large language models with minimal setup. KleidiAI is an Arm open-source library that provides optimized micro-kernels to accelerate AI workloads on Arm CPU architectures.