The ggml‑org/llama.cpp project published version b11641 on 15 Oct 2026. The update adds lilicorr support, a new feature for handling certain correlation calculations. It also includes minor typo fixes and updates to hyper‑parameter handling code. The release lists builds for macOS, Linux, Android, Windows and several GPU back‑ends such as CUDA 12/13, Vulkan and ROCm. The project's website is llama.app.
Why it matters
Developers can now run llama.cpp with lilicorr on more hardware, enabling slightly more efficient inference for supported models.
