Total runs:
357.8K
Run Growth:
-283.7K
Growth Rate:
-79.28%
Sponsored by
Vmake Labs Video Enhancer
z-lab / Qwen3.5-27B-DFlash
Model's Last Updated: June 20 2026
text-generation
Introduction of Qwen3.5-27B-DFlash
Model Details of Qwen3.5-27B-DFlash
Qwen3.5-27B-DFlash
This model is still under training.
DFlash is a novel speculative decoding method that utilizes a lightweight block diffusion model for drafting. It enables efficient, high-quality parallel drafting that pushes the limits of inference speed.
This model is the
drafter
component. It must be used in conjunction with the target model
Qwen/Qwen3.5-27B
. It was trained with a context length of 4096 tokens.
🚀 Quick Start
SGLang
Installation
uv pip install "git+https://github.com/sgl-project/sglang.git@refs/pull/16818/head#subdirectory=python"
Inference
python -m sglang.launch_server \
--model-path Qwen/Qwen3.5-27B \
--speculative-algorithm DFLASH \
--speculative-draft-model-path z-lab/Qwen3.5-27B-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
Note: For long-context or agentic usage, consider adding
--speculative-dflash-draft-window-size WINDOW_SIZEto enable sliding-window attention for the draft model.
vLLM
Thanks to the community and all contributors! Check out the following PRs to see how to run DFlash on vLLM: #36847 and #36767 .
Early Results
- Thinking: enabled
- Max new tokens: 4096
-
Block size: 16
Dataset Accept Length GSM8K xxx Math500 xxx HumanEval xxx MBPP xxx MT-Bench xxx Alpaca xxx
Runs of z-lab Qwen3.5-27B-DFlash on huggingface.co
3.5K
Total runs
-53
24-hour runs
-141
3-day runs
-512
7-day runs
-3.0K
30-day runs
More Information About Qwen3.5-27B-DFlash huggingface.co Model
Url of Qwen3.5-27B-DFlash
Qwen3.5-27B-DFlash huggingface.co Url
Provider of Qwen3.5-27B-DFlash huggingface.co
Other API from z-lab
Updated:August 25 2026
Total runs:
96.4K
Run Growth:
-265.2K
Growth Rate:
-275.14%
Total runs:
48.8K
Run Growth:
33.7K
Growth Rate:
69.07%
Total runs:
45.7K
Run Growth:
-90.3K
Growth Rate:
-197.85%
Total runs:
28.1K
Run Growth:
23.0K
Growth Rate:
81.86%
Total runs:
27.2K
Run Growth:
23.1K
Growth Rate:
84.69%
Total runs:
26.5K
Run Growth:
-140.8K
Growth Rate:
-531.96%
Total runs:
16.3K
Run Growth:
1.5K
Growth Rate:
9.37%
Total runs:
16.3K
Run Growth:
-2.8K
Growth Rate:
-17.05%
Total runs:
13.3K
Run Growth:
-9.8K
Growth Rate:
-73.37%
Total runs:
5.9K
Run Growth:
-6.5K
Growth Rate:
-109.66%
Total runs:
5.1K
Run Growth:
-24.5K
Growth Rate:
-483.68%
Total runs:
3.4K
Run Growth:
-4.6K
Growth Rate:
-134.81%
Total runs:
2.9K
Run Growth:
-1.2K
Growth Rate:
-41.10%
Total runs:
2.4K
Run Growth:
-4.4K
Growth Rate:
-181.81%
Total runs:
1.5K
Run Growth:
-1.1K
Growth Rate:
-70.90%
Total runs:
1.4K
Run Growth:
1.2K
Growth Rate:
87.09%
Total runs:
1.2K
Run Growth:
237
Growth Rate:
19.10%
Total runs:
1.2K
Run Growth:
1.2K
Growth Rate:
100.00%
Total runs:
1.1K
Run Growth:
-4.8K
Growth Rate:
-430.15%
Total runs:
1.1K
Run Growth:
-6.3K
Growth Rate:
-580.20%
Total runs:
777
Run Growth:
61
Growth Rate:
7.85%
Total runs:
435
Run Growth:
-1.5K
Growth Rate:
-340.92%
Total runs:
331
Run Growth:
101
Growth Rate:
30.51%
Total runs:
296
Run Growth:
-1.4K
Growth Rate:
-469.93%
Total runs:
141
Run Growth:
47
Growth Rate:
33.33%
Total runs:
104
Run Growth:
77
Growth Rate:
74.04%
Total runs:
80
Run Growth:
0
Growth Rate:
0.00%
Total runs:
77
Run Growth:
-17
Growth Rate:
-22.08%
Total runs:
49
Run Growth:
24
Growth Rate:
48.98%
Total runs:
30
Run Growth:
1
Growth Rate:
3.33%
Total runs:
26
Run Growth:
10
Growth Rate:
38.46%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%