The ggml‑org project released llama.cpp version b11772 on Oct 21, 2026. It includes a fix for the Gemma‑4 feed‑forward network activation. The release bundles pre‑compiled binaries for macOS Apple Silicon, Intel, iOS, and a range of Ubuntu CPUs, Vulkan and CUDA‑12.8 GPUs.
Developers can download the appropriate tar.gz or zip files from the GitHub release page and run LLMs locally without building from source.
Why it matters
It lets developers run Gemma‑4 models on more hardware with fewer crashes.
