inferRouter/Qwen3.6-27B-FP8-lmhead-embed-fp8

docs(model card): RTX 5090 16K production verdict — quality-validated within ≤10K prompt + 6144 completion envelope; 8% coverage caveat; link ADR-0129

docs(model card): drop standalone PR #41000 from overlay list (carried by stacked #41365); add production ceiling profile; add inference: false; cross-link lmhead-fp8 baseline for eval

Upload Qwen3.6-27B FP8 lm_head+embed_tokens artifact

initial commit