Local Ai
b9106
Release b9106 of llama.cpp includes support for asymmetric FlashAttention in scalar, MMQ, and cooperative matrix paths for Vulkan. The release provides builds for multiple platforms including macOS, L
Release b9106 of llama.cpp includes support for asymmetric FlashAttention in scalar, MMQ, and cooperative matrix paths for Vulkan. The release provides builds for multiple platforms including macOS, Linux, Windows, and Android with support for various accelerators such as Vulkan, ROCm, CUDA, and OpenVINO.
Source: llama.cpp Releases | 2026-05-11