This guide shows how to run NVIDIA's optimized Qwen3.6-35B-A3B-NVFP4 model on an NVIDIA DGX Spark using vLLM.
The result is a local, OpenAI-compatible API endpoint that can be used by applications such as Open WebUI, OpenClaw, LangChain, the OpenAI Python SDK, or any client that supports the OpenAI chat-completions API.
When complete, the endpoint will be available at:
http://:8000/v1