The open‑source llama.cpp library released version b11706 on 2026-10-17. The update fixes a regression in SYCL prompt processing that appeared after weight reordering for Q4_0/Q8_0 formats. It also ships new pre‑built binaries for macOS Apple Silicon, Intel, iOS, and various Linux configurations including CPU, Vulkan and CUDA 12.8. The builds are available for download from the GitHub release page.
Why it matters
Developers can now run llama.cpp on SYCL‑compatible GPUs without prompt‑handling errors, expanding hardware options for local LLM use.
