The ggml-org team has released llama.cpp b11345, adding support for the q2_k and q3_k quantization types on Hexagon. Quantization is a way to store model data using fewer bits; the release notes do not say which devices or models benefit, or describe performance changes.
The release also includes downloads for macOS, iOS and several Linux configurations, including CPU and Vulkan builds.
Why it matters
People running llama.cpp on supported Hexagon hardware may be able to use models with q2_k or q3_k quantization, but the release notes do not specify device compatibility or performance effects.
