The ggml-org team released llama.cpp b12368 on November 13, 2026. Its listed change improves buffer allocation and memory mapping for Hexagon, including retries for mapping failures and changes intended to reduce memory-address fragmentation.
The release page also offers builds for macOS, iOS and several Linux systems. It does not describe performance gains or specify which devices benefit from the Hexagon changes.
Why it matters
Developers using llama.cpp with Hexagon hardware get updated memory-handling code, but the release notes do not say whether it improves speed or reliability in practice.
