Running OpenClaw on a DGX Spark with a local Qwen model
From the community · not tested by us
They run OpenClaw on an NVIDIA DGX Spark with a local Qwen model via vLLM at about 93 tokens a second and a 262K context, and wrote up the setup.
Details
Posted by@vaishaksalin
DateOct 18, 2026
Category⚙️ Automation & ops
Found viaX
Summary written by ToolPassed from the public post. Claims are the poster’s; we haven’t reproduced this.