Qwen3.8 27B qwen3.8-27b-uncensored

Your API key stays in this browser. Requests go directly to the Spark over Tailscale.

DEPLOYMENT

DGX Spark status

Runtime configuration and measured validation from this machine.

MODEL27B FP8Qwen3.8 uncensored
CONTEXT32KInitial window
OUTPUT16.264Median tok/s
MEMORY45%vLLM target
Runtime pathLIVE
BROWSERPages UI
→
PRIVATE TLSTailscale
→
LOCAL APIvLLM :8100
→
ACCELERATORGB10
Validated capabilities10 / 10
  • ● Model integrity
  • ● Authenticated health
  • ● Deterministic chat
  • ● Coherent output
  • ● SSE streaming
  • ● Structured tools
  • ● Image input
  • ● Tailnet HTTPS
  • ● Secret isolation
  • ● LTX isolation
Pinned deployment2026-08-20
Model revision
9228df5c6c9c509e1019f83b4e085cf643118bac
Runtime
NVIDIA vLLM 26.07 · vLLM 0.24 · XGrammar 0.2.1
Private endpoint
spark-2944-dgx-userspace.tailc87c5a.ts.net:8443/v1

CONNECTION

Qwen settings