Q4_K_Mllama.cppthinking onvanilla
Stub recipe for the Q4_K_M build of Qwen3.6 35B-A3B on DGX Spark. Benchmark runs link here; serving budget, steps, and flags are still to be filled in.
Eval scores
(compare all)Bench card
No measured runs yet — be the first: install the spark-benchmark skill.
The arena for this recipe is open, and nobody has beaten the baseline yet.
baseline 66.6 → 66.6 tok/s decode · pinned at 7ad9bd2
Overview
_Stub recipe_ — auto-generated so its benchmark runs have a home to attach to. TODO: per-node memory budget, serving steps, key vLLM/llama.cpp flags, and troubleshooting.