Projects

ggml-org/llama.cpp

LLM inference in C/C++

Posts

ReleaseOCT 17

llama.cpp b11700 added decision conversion and support paths for Granite Switch family models.

ReleaseOCT 16

llama.cpp nightly b11683 is a build that fixes the Vulkan backend's host memory import failure logging and AMD's buffer sharing path.

ReleaseOCT 16

ggml-org/llama.cpp b11630 is a daily build automated release whose changes are described only at the level of commit messages.