The ggml‑org team published llama.cpp version b11790 on Oct 21 2026. The update adds neon‑vectorized repack for Q4_0 quantization and threads other types, improving performance on ARM CPUs. Pre‑built binaries are now available for macOS Apple Silicon, iOS, and various Linux ARM builds. The changes were co‑authored by a Qualcomm engineer. The release also includes Vulkan builds for Linux x64 and arm64.
Why it matters
ARM‑based laptops and phones can run Llama models faster, making local AI use more practical.
