The ggml‑org team released llama.cpp version b11792 on 2026‑10‑21. The update changes the set_rows call so data can stay on the GPU, improving speed for GPU users. Pre‑built binaries are now available for macOS Apple Silicon, macOS Intel, iOS, and several Ubuntu builds (CPU, Vulkan, CUDA 12.8). The release links are posted on the GitHub tag page.
The change is listed in the release notes as “place d2t by its set_rows use so it can stay on the GPU (#30912)”.
Why it matters
Developers can run Llama models faster on compatible GPUs without extra data transfers.
