| --- |
| license: mit |
| library_name: transformers |
| pipeline_tag: text-generation |
| tags: |
| - dflash |
| - speculative-decoding |
| - block-diffusion |
| - draft-model |
| - efficiency |
| - qwen |
| - diffusion-language-model |
| --- |
| |
| # Qwen3-Coder-Next-DFlash |
|
|
| [**Paper**](https://arxiv.org/abs/2602.06036) | [**GitHub**](https://github.com/z-lab/dflash) | [**Blog**](https://z-lab.ai/projects/dflash/) |
|
|
| **DFlash** is a speculative decoding method that uses a lightweight **block diffusion** model to draft multiple tokens in parallel. This is the drafter model, which must be paired with [Qwen/Qwen3-Coder-Next](https://huggingface.co/Qwen/Qwen3-Coder-Next). |
|
|
| <div align="center"> |
| <img src="assets/dflash_system.png" alt="DFlash Architecture" width="85%"> |
| </div> |
|
|
| ## Quick Start |
|
|
| ### Installation |
|
|
| ```bash |
| uv pip install "git+https://github.com/sgl-project/sglang.git@refs/pull/20547/head#subdirectory=python" |
| ``` |
|
|
| ### Launch Server |
|
|
| Use `--speculative-num-draft-tokens` to set the block size (8 or **16**). |
|
|
| ```bash |
| export SGLANG_ENABLE_SPEC_V2=1 |
| export SGLANG_ENABLE_DFLASH_SPEC_V2=1 |
| export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1 |
| |
| python -m sglang.launch_server \ |
| --model-path Qwen/Qwen3-Coder-Next \ |
| --speculative-algorithm DFLASH \ |
| --speculative-draft-model-path z-lab/Qwen3-Coder-Next-DFlash \ |
| --speculative-num-draft-tokens 16 \ |
| --tp-size 1 \ |
| --attention-backend fa3 \ |
| --mem-fraction-static 0.75 \ |
| --mamba-scheduler-strategy extra_buffer \ |
| --trust-remote-code |
| ``` |
|
|
| > **Tip:** For long-context or agentic workloads, add `--speculative-dflash-draft-window-size WINDOW_SIZE` to enable sliding-window attention for the drafter. |
| |
| ### Usage |
| |
| ```python |
| from openai import OpenAI |
| |
| client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY") |
| |
| response = client.chat.completions.create( |
| model="Qwen/Qwen3-Coder-Next", |
| messages=[{"role": "user", "content": "Write a quicksort in Python."}], |
| max_tokens=4096, |
| ) |
| print(response.choices[0].message.content) |
| ``` |
| |
| ### vLLM |
| |
| Community-contributed support is available. See PRs [#36847](https://github.com/vllm-project/vllm/pull/36847) and [#36767](https://github.com/vllm-project/vllm/pull/36767) for details. |
| |
| ## Acceptance Length |
| |
| - Max new tokens: 4096 |
| - Block size: 16 |
| | Dataset | Accept Length | |
| |-----------|---------------| |
| | HumanEval | 7.25 | |
| | MBPP | 5.50 | |
| | LiveCodeBench | 5.50 | |
| |
| ## Acknowledgements |
| |
| Special thanks to [David Wang](https://davidwa.ng/) for his outstanding engineering support on this project. We are also grateful to [Modal](https://modal.com/), [InnoMatrix](https://innomatrix.ai), and [Yotta Labs](https://www.yottalabs.ai/) for providing the compute resources used to train this draft model. |
| |
| ## Citation |
| |
| If you find DFlash useful, please cite our work. To share feedback on DFlash or request new model support, please fill out this form: [DFlash Feedback](https://forms.gle/4YNwfqb4nJdqn6hq9). |
| |
| ```bibtex |
| @article{chen2026dflash, |
| title = {{DFlash: Block Diffusion for Flash Speculative Decoding}}, |
| author = {Chen, Jian and Liang, Yesheng and Liu, Zhijian}, |
| journal = {arXiv preprint arXiv:2602.06036}, |
| year = {2026} |
| } |
| ``` |
