DFlash: Block Diffusion for Flash Speculative Decoding

Paper | Blog | Models

DFlash is a lightweight block diffusion model designed for speculative decoding. It enables efficient and high-quality parallel drafting.

DFlash_demo.mp4

📦 Model Support Plan

✅ Supported

openai/gpt-oss-20b: https://huggingface.co/z-lab/gpt-oss-20b-DFlash
Qwen3-4B: https://huggingface.co/z-lab/Qwen3-4B-DFlash-b16
Qwen3-8B: https://huggingface.co/z-lab/Qwen3-8B-DFlash-b16
Qwen3-Coder-30B-A3B: https://huggingface.co/z-lab/Qwen3-Coder-30B-A3B-DFlash
Llama-3.1-8B-Instruct: https://huggingface.co/z-lab/LLaMA3.1-8B-Instruct-DFlash-UltraChat

🚧 Coming Soon

Qwen/Qwen3-Coder-Next (Very soon)
openai/gpt-oss-120b
zai-org/GLM-4.7
zai-org/GLM-4.7-Flash

💡 Feel free to open a GitHub issue if you’d like to request support for additional models!
We will also open-source the training recipe soon, so you can train your own DFlash draft model to accelerate any LLM.

🚀 Quick Start

Installation

conda create -n dflash python=3.11
conda activate dflash

git clone https://github.com/z-lab/dflash.git
cd dflash

pip install uv
uv pip install -r requirements.txt

# Optionally install flash-attn.
# If unavailable, evaluation falls back to torch.sdpa in the Transformers backend.
# The measured speedup will be slower, but the acceptance length remains comparable.

# uv pip install flash-attn --no-build-isolation

SGLang

export SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1

python -m sglang.launch_server \
    --model-path Qwen/Qwen3-Coder-30B-A3B-Instruct \
    --speculative-algorithm DFLASH \
    --speculative-draft-model-path z-lab/Qwen3-Coder-30B-A3B-DFlash \
    --tp-size 1 \
    --dtype bfloat16 \
    --attention-backend fa3 \
    --mem-fraction-static 0.75 \
    --trust-remote-code

Transformers

from transformers import AutoModel, AutoModelForCausalLM, AutoTokenizer

model = AutoModel.from_pretrained(
    "z-lab/Qwen3-8B-DFlash-b16", 
    trust_remote_code=True, 
    dtype="auto", 
    device_map="cuda:0"
).eval()

target = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen3-8B", 
    dtype="auto", 
    device_map="cuda:0"
).eval()

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-8B")
prompt = "How many positive whole-number divisors does 196 have?"
messages = [
    {"role": "user", "content": prompt}
]
# Note: this draft model is used for thinking mode disabled
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=False
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generate_ids = model.spec_generate(
    input_ids=model_inputs["input_ids"], 
    max_new_tokens=2048, 
    temperature=0.0, 
    target=target, 
    stop_token_ids=[tokenizer.eos_token_id]
)

print(tokenizer.decode(generate_ids[0], skip_special_tokens=False))

📊 Evaluation

We provide scripts to reproduce the speedup and acceptance length metrics in the paper. The reported results were tested on NVIDIA H200 or B200 GPUs.

To run benchmark on Transformers backend:

bash run_benchmark.sh

To run benchmark on SGLang:

export SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1

python benchmark_sglang.py \
  --target-model Qwen/Qwen3-8B \
  --draft-model z-lab/Qwen3-8B-DFlash-b16 \
  --concurrencies 1,4,8,16,32 \
  --dataset-name math500 \
  --attention-backends fa3,flashinfer \
  --tp-size 1 \
  --output-md sglang_results.md

Acknowledgement

Huge thanks to @dcw02, @gongy, and the other folks at @modal-labs for the fast, high-quality support in bringing DFlash into SGLang—making it possible to truly accelerate LLM serving in real-world deployments.

Citation

If you find DFlash useful for your research or applications, please cite our project.

@misc{chen2026dflash,
  title         = {DFlash: Block Diffusion for Flash Speculative Decoding},
  author        = {Chen, Jian and Liang, Yesheng and Liu, Zhijian},
  year          = {2026},
  eprint        = {2602.06036},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CL},
  url           = {https://arxiv.org/abs/2602.06036}
}

zhaohb/dflash