z-lab/Qwen3.8-27B-DFlash2 · Hugging Face

Source: https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2 Blog|GitHub

This repository contains the DFlash 2 draft model forQwen/Qwen3\.8\-27B. It is not a standalone language model: it runs inside a speculative decoding server and drafts tokens for the target model to verify. This repository is a mirror ofincoai/Qwen3\.8\-27B\-DFlash2.

DFlash 2 is a block-diffusion drafter for speculative decoding. It predicts a whole block of tokens in a single pass and keeps the top candidates at every position. A lightweight selector then traces one coherent path through them. Two-tap dynamic convolutions in the backbone keep the draft from decaying toward the end of the block. Decoding is lossless: greedy output matches the target model exactly, and sampling preserves its distribution.

https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2#quick-startQuick Start

Serve withSGLang:

pip install "sglang[all] @ git+https://github.com/sgl-project/sglang.git#subdirectory=python"

python -m sglang.launch_server \
  --model-path Qwen/Qwen3.8-27B \
  --speculative-algorithm DFLASH \
  --speculative-draft-model-path incoai/Qwen3.8-27B-DFlash2 \
  --speculative-num-draft-tokens 8

Or withvLLM:

pip install -U "vllm @ git+https://github.com/vllm-project/vllm.git@refs/pull/52816/head"

vllm serve Qwen/Qwen3.8-27B \
  --speculative-config '{
    "method": "dflash",
    "model": "incoai/Qwen3.8-27B-DFlash2",
    "num_speculative_tokens": 7
  }'

See theblog postfor other engines and more details.

https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2#evaluationEvaluation

• Runtime: SGLang on one NVIDIA H200, with FlashAttention 3 for target and draft attention
• Speculation block size: 8 (7 draft tokens per verification step)
• Sampling: Qwen3.8’s officially recommended parameters (temperature 1.0, top-p 0.95, top-k 20), withxhighreasoning effort
• Maximum new tokens: 4096
• Prompts: benchmark formatting fromz\-lab/dflash

We compare autoregressive decoding, Qwen3.8’s built-in seven-token MTP, a community DSpark drafter (RadixArk/Qwen3\.8\-27B\-DSpark), and DFlash 2. All speculative methods propose seven draft tokens per verification step.

https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2#acceptance-lengthAcceptance Length

Acceptance length is the per-request mean of completion tokens divided by verification steps. Higher is better.

TaskMTPDSparkDFlash 2GSM8K5.024.365.46MATH-5004.723.925.28HumanEval3.913.304.39MBPP3.993.514.79MT-Bench3.743.014.10

https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2#throughputThroughput

Throughput is total output tokens divided by end-to-end wall time. Each cell showsoutput tok/s \(speedup vs\. autoregressive\).

https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2#concurrency-1Concurrency 1

TaskAutoregressiveMTPDSparkDFlash 2GSM8K68.9178.5 (2.59×)185.3 (2.69×)**236.1 (3.43×)**MATH-50069.0172.8 (2.51×)174.5 (2.53×)**230.7 (3.34×)**HumanEval69.0151.9 (2.20×)159.9 (2.32×)**214.6 (3.11×)**MBPP69.0153.1 (2.22×)163.3 (2.37×)**226.9 (3.29×)**MT-Bench68.9134.9 (1.96×)137.6 (2.00×)184.0 (2.67×)

https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2#concurrency-8Concurrency 8

TaskAutoregressiveMTPDSparkDFlash 2GSM8K467.21,022.1 (2.19×)1,040.8 (2.23×)**1,328.7 (2.84×)**MATH-500480.01,023.5 (2.13×)1,025.8 (2.14×)**1,368.3 (2.85×)**HumanEval483.4934.2 (1.93×)956.5 (1.98×)**1,291.5 (2.67×)**MBPP478.0938.1 (1.96×)974.1 (2.04×)**1,328.0 (2.78×)**MT-Bench480.5835.2 (1.74×)802.3 (1.67×)1,090.2 (2.27×)

https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2#concurrency-32Concurrency 32

TaskAutoregressiveMTPDSparkDFlash 2GSM8K1,329.81,381.1 (1.04×)1,506.5 (1.13×)**1,922.5 (1.45×)**MATH-5001,505.81,415.6 (0.94×)1,429.0 (0.95×)**1,951.8 (1.30×)**HumanEval1,546.51,296.8 (0.84×)1,330.1 (0.86×)**1,799.0 (1.16×)**MBPP1,507.71,314.9 (0.87×)1,361.3 (0.90×)**1,886.8 (1.25×)**MT-Bench1,507.41,159.7 (0.77×)1,115.5 (0.74×)1,525.3 (1.01×)

https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2#citationCitation

If you find DFlash 2 useful, please cite:

@misc{inco2026dflash2,
  title  = {{DFlash 2: Keep Drafting Parallel}},
  author = {{Inco AI}},
  year   = {2026},
  month  = {August},
  url    = {https://inco.ai/blog/dflash2/}
}

Please also cite the original DFlash paper:

@inproceedings{chen2026dflash,
  title     = {{DFlash: Block Diffusion for Flash Speculative Decoding}},
  author    = {Chen, Jian and Liang, Yesheng and Liu, Zhijian},
  booktitle = {International Conference on Machine Learning (ICML)},
  year      = {2026}
}