Full threadw-m·GGUF support was merged in llama.cpp three days ago: https://github.com/ggerganov/llama.cpp/pull/2398View on HN