---
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
tags:
- dflash
- speculative-decoding
- block-diffusion
- draft-model
- efficiency
- qwen
- gemma
- diffusion-language-model
---

# gemma-4-26B-A4B-it-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 [google/gemma-4-26B-A4B-it](https://huggingface.co/google/gemma-4-26B-A4B-it).

<div align="center">
  <img src="assets/dflash_system.png" alt="DFlash Architecture" width="85%">
</div>

## Quick Start

### Installation

vLLM: until Gemma4 DFlash support is merged, install vLLM from [PR #41703](https://github.com/vllm-project/vllm/pull/41703):

```bash
uv pip install -U --torch-backend=auto \
  "vllm @ git+https://github.com/vllm-project/vllm.git@refs/pull/41703/head"
```

SGLang:

```bash
uv pip install "git+https://github.com/sgl-project/sglang.git@refs/pull/23000/head#subdirectory=python"
```

### Launch Server

vLLM:

```bash
vllm serve google/gemma-4-26B-A4B-it \
  --speculative-config '{"method": "dflash", "model": "z-lab/gemma-4-26B-A4B-it-DFlash", "num_speculative_tokens": 15, "attention_backend": "flash_attn"}' \
  --attention-backend triton_attn \
  --max-num-batched-tokens 32768 \
  --trust-remote-code
```

SGLang:

```bash
# Optional: enable schedule overlapping (experimental, may not be stable)
# 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 google/gemma-4-26B-A4B-it \
  --speculative-algorithm DFLASH \
  --speculative-draft-model-path z-lab/gemma-4-26B-A4B-it-DFlash \
  --speculative-num-draft-tokens 16 \
  --tp-size 1 \
  --attention-backend triton \
  --speculative-draft-attention-backend fa4 \
  --trust-remote-code
```

### Usage

For vLLM, use port `8000`. For SGLang, use port `30000`.

```python
from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")

response = client.chat.completions.create(
    model="google/gemma-4-26B-A4B-it",
    messages=[{"role": "user", "content": "Write a quicksort in Python."}],
    max_tokens=4096,
    temperature=0.0,
    extra_body={"chat_template_kwargs": {"enable_thinking": True}},
)
print(response.choices[0].message.content)
```

## Benchmark Results

**Setup:** Single NVIDIA B300 GPU per server/run, vLLM, thinking enabled, max output length 4096, greedy decoding.

### Throughput and Speedup

DFlash achieves up to **3.7x** speedup at concurrency 8.

_Generated tokens/sec (speedup vs. autoregressive baseline)_

**Block Size = 16**

| Task | Concurrency | AR | **DFlash** |
|---|---:|---:|---:|
| Math500 | 1 | 259 | **925 (3.6x)** |
|  | 8 | 1296 | **4837 (3.7x)** |
|  | 32 | 3233 | **11435 (3.5x)** |
| GSM8K | 1 | 256 | **825 (3.2x)** |
|  | 8 | 1217 | **4241 (3.5x)** |
|  | 32 | 3174 | **10306 (3.2x)** |
| HumanEval | 1 | 246 | **818 (3.3x)** |
|  | 8 | 1182 | **4240 (3.6x)** |
|  | 32 | 2881 | **9150 (3.2x)** |
| MBPP | 1 | 272 | **698 (2.6x)** |
|  | 8 | 1288 | **3387 (2.6x)** |
|  | 32 | 2950 | **7898 (2.7x)** |
| MT-Bench | 1 | 272 | **492 (1.8x)** |
|  | 8 | 1146 | **2259 (2.0x)** |
|  | 32 | 2164 | **4829 (2.2x)** |

### Acceptance Length

| Task | c1 | c8 | c32 |
|---|---:|---:|---:|
| Math500 | 8.61 | 8.55 | 8.60 |
| GSM8K | 7.71 | 7.76 | 7.72 |
| HumanEval | 7.80 | 7.87 | 7.83 |
| MBPP | 6.09 | 5.99 | 6.03 |
| MT-Bench | 4.33 | 4.33 | 4.24 |

## 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}
}
```