5d ago · llama.cpp releases →
mtmd: refactoring: remove clip.h, consolidate code into mtmd (#31865)
Includes: llama.cpp b12206, b12207: ggml-cuda : add d_state=64 fast path to SSM_SCAN kernel (#31818), llama.cpp b12208, b12209: llamafile: impl bf16 and q8_0 vxe (#31866), b12210: cpu: enable SpacemiT IME2 MXFP4 dense repack (#31776), llama.cpp b12211