Qwen3.6-27B (dense) · DDTree (Diffusion Draft Tree) · BLOCKED (hybrid GDN cache)

status: blocked

Configuration

ModelQwen/Qwen3.6-27B
CompanyAlibaba
FamilyQwen
Parameters27B dense (hybrid GatedDeltaNet + full-attn)
EngineDDTree research harness (github.com/liranringel/ddtree, PyTorch + transformers, batch-1) + DDTree — tree draft over the z-lab/Qwen3.6-27B-DFlash block-diffusion drafter (speculative decoding)
Quant / precisionBF16 (harness loads the unquantized target via AutoModelForCausalLM)
Why this quantEXPERIMENTS.md P0
DownloadQwen/Qwen3.6-27B
Context window4096
Input modalitiestext (served text-only here)

Measured results

Prefill tok/s
Decode tok/s
Peak memory (GB)
Completed2026-07-01

Full run command

scripts/bench-ddtree.sh Qwen/Qwen3.6-27B z-lab/Qwen3.6-27B-DFlash mt-bench 20 512 64,256 0.0 q27b
# Same GatedDeltaNet rollback wall as the 35B-A3B.

qwen3.6-27b Alibaba Qwen BF16 16-40B Spark recipe conc-1

Notes

BLOCKED — same hybrid GatedDeltaNet wall as qwen3-6-35b-a3b-ddtree-blocked. Qwen3.6-27B is model_type: qwen3_5 (Qwen3_5ForConditionalGeneration) — the dense member of the same Qwen3.5/3.6 family, with the same interleaved GatedDeltaNet linear-attention + full-attention stack. The DDTree/DFlash harness verifies speculative blocks by past_key_values.crop(start) after each block, which a recurrent linear-attention state cannot support (nothing to slice back to an arbitrary accepted position), so the target forward fails identically. No serving engine implements DDTree, and the harness has no hybrid-cache/state-checkpoint path.

The measurable DDTree datapoint on the Spark is on the harness-supported non-hybrid target — see qwen3-coder-30b-a3b-ddtree.