Downloaded IQ4_NL gguf from unsloth/Qwen3.6-35B-A3B-MTP-GGUF. git cloned a recent llama.cpp (version: 9397 (ac4b5a3fd)) and compiled it with GGML_CUDA=ON to run on my single 3090 llama-server command without MTP: ./build/bin/llama-server -m ~/gguf/Qwen3.6-35B-A3B-UD-IQ4_NL.gguf --host 0.0.0.0 --port
