IK_LLAMA now supports Qwen3.5 MTP Support :O
Compile, compile, compile! https://github.com/ikawrakow/ik_llama.cpp/pull/1698 Will be testing shortly! EDIT: You will need a GGUF with the MTP layers preserved. The PR creator made some GGUFs of Q3.6 27B at Q8_0 here – https://huggingface.co/Radamanth…