llama.cpp

revert : "[Model] Qwen3.5 dense and MoE support (no vision) (#19435)"

#19453

Merged