Lemonade GPU deployment: Qwen3.6-35B-A3B

Run Qwen3.6 GGUF on the iGPU through Lemonade llamacpp.

Prerequisite: Lemonade environment. Model: Qwen3.6. NPU: FastFlowLM.

lemonade backends install llamacpp:vulkan
lemonade pull Qwen3.6-35B-A3B-GGUF --source modelscope
lemonade load Qwen3.6-35B-A3B-GGUF --llamacpp vulkan
lemonade status

Id must contain 3.6. Use Vulkan if llamacpp:rocm is unsupported. Smoke with BASE=http://127.0.0.1:13305/api/v1 MODEL=Qwen3.6-35B-A3B-GGUF python3 src/npu/examples/chat_gemma4.py. Do not keep this resident together with the ~29 GiB NPU model on 8219.