---
title: Qwen3.8-27B-DSpine-b16
canonical_url: "https://www.modelscope.ai/models/tencent-community/Qwen3.8-27B-DSpine-b16"
md_url: "https://www.modelscope.ai/models/tencent-community/Qwen3.8-27B-DSpine-b16.md"
repository: tencent-community/Qwen3.8-27B-DSpine-b16
last_updated: 2026-11-10
license: apache-2.0
pipeline_tag: text-generation
tasks:
- text-generation
model_type:
- qwen3
architectures:
- DFlashDraftModel
base_model:
- Qwen/Qwen3.8-27B
base_model_relation: finetune
parameters: 2.0B
tensor_type:
- BF16
- I64
library_name:
- safetensors
- pytorch
frameworks:
- pytorch
downloads: 11
stars: 0
tags:
- dspine
- speculative-decoding
- draft-model
- sglang
---
# Qwen3.8-27B-DSpine-b16
> Qwen3.8-27B-DSpine-b16 - An open-source model by tencent-community on ModelScope. DSpine for Qwen3.8-27B
tencent-community/Qwen3.8-27B-DSpine-b16 is a 2.0B-parameter text-generation model on ModelScope. licensed under apache-2.0. derived from Qwen/Qwen3.8-27B.
- **Repository**: tencent-community/Qwen3.8-27B-DSpine-b16
- **License**: apache-2.0
- **Tasks**: text-generation
- **Parameters**: 2.0B
- **Base model**: Qwen/Qwen3.8-27B
- **Tags**: dspine, speculative-decoding, draft-model, sglang
- **Downloads**: 11
- **Stars**: 0
- **Last updated**: 2026-11-10
Source: https://www.modelscope.ai/models/tencent-community/Qwen3.8-27B-DSpine-b16
---
DSpine for Qwen3.8-27B
DSpine generates draft tokens in parallel. Each layer passes predicted predecessor features to adjacent positions, carrying causal dependencies through network depth.
This checkpoint is a five-layer draft model for [Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B), with block size 16 and an 8,192-token sliding window.
## Evaluation
- **Runtime:** SGLang 0.5.22, BF16, FlashAttention 3, concurrency 8.
- **Draft blocks:** DSpine 16; official [DFlash2](https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2) 8.
- **DSpine inference:** dynamic verification referencing [DSpark](https://github.com/deepseek-ai/DeepSpec), with [ReplaySSM](https://github.com/Johnny-Liou/ReplaySSM) enabled.
- **General benchmarks:** `xhigh` reasoning, greedy and temperature sampling, at most 4,096 new tokens.
- **Agent benchmarks:** We evaluate [WorkBuddy Bench](https://github.com/Tencent/workbuddy-bench) (Code/Web/Office) using CodeBuddy sandbox trajectories: `low`/`xhigh` reasoning, temperature sampling, 20 conversations per subset, all tool-call prefixes replayed in order, prefix caching, and up to 32,768 new tokens per request.
- **Temperature sampling:** `temperature=1.0`, `top_p=0.95`, `top_k=20`.
## Results
### General Benchmark Acceptance Length
| Task | DFlash2(T=0) | DSpine(T=0) | DFlash2(T=1) | DSpine(T=1) |
| --- | ---: | ---: | ---: | ---: |
| GSM8K | 5.69 | **7.20** | 5.45 | **6.31** |
| MATH-500 | 5.49 | **6.49** | 5.27 | **5.66** |
| HumanEval | 4.87 | **5.71** | 4.34 | **4.71** |
| MBPP | 5.04 | **6.14** | 4.79 | **5.41** |
| MT-Bench | 4.11 | **4.48** | 3.89 | **4.03** |
### Agent Acceptance Length
| Benchmark | low | xhigh |
|---|
| DFlash2 | DSpine | DFlash2 | DSpine |
|---|
| WorkBuddy Code | 4.19 | 5.08 | 4.06 | 4.92 |
| WorkBuddy Web | 3.95 | 4.91 | 3.95 | 5.01 |
| WorkBuddy Office | 4.24 | 5.63 | 4.20 | 5.49 |
| SWE-bench Pro | 4.19 | 4.99 | 4.00 | 4.77 |
| τ²-bench | 4.70 | 5.53 | 4.29 | 4.95 |
### Agent End-to-End Throughput
Throughput is total output tokens divided by end-to-end concurrent replay wall time (tokens/s). Parentheses show speedup over autoregressive decoding (AR).
| Benchmark | low | xhigh |
|---|
| AR | DFlash2 | DSpine | AR | DFlash2 | DSpine |
|---|
| WorkBuddy Code | 253.35 | 590.27 (2.33×) | 686.23 (2.71×) | 193.00 | 550.40 (2.85×) | 613.55 (3.18×) |
| WorkBuddy Web | 247.40 | 639.74 (2.59×) | 773.19 (3.13×) | 252.26 | 734.04 (2.91×) | 840.14 (3.33×) |
| WorkBuddy Office | 210.03 | 664.77 (3.17×) | 805.53 (3.84×) | 215.05 | 662.34 (3.08×) | 763.83 (3.55×) |
| SWE-bench Pro | 241.86 | 657.71 (2.72×) | 713.41 (2.95×) | 248.17 | 711.60 (2.87×) | 791.30 (3.19×) |
| τ²-bench | 301.39 | 891.37 (2.96×) | 964.30 (3.20×) | 340.39 | 871.98 (2.56×) | 978.39 (2.87×) |
## Installation
```bash
git clone https://github.com/Tencent/DSpine.git
cd DSpine
pip install -e .
```
## Usage
```bash
sglang serve \
--model-path Qwen/Qwen3.8-27B \
--speculative-algorithm DSPINE \
--speculative-draft-model-path /path/to/Qwen3.8-27B-DSpine-b16 \
--speculative-num-draft-tokens 16 \
--dtype bfloat16 \
--attention-backend fa3 \
--speculative-draft-attention-backend fa3 \
--linear-attn-backend triton \
--enable-linear-replayssm-spec \
--mamba-ssm-dtype float32 \
--mamba-radix-cache-strategy extra_buffer \
--max-mamba-cache-size 64 \
--mem-fraction-static 0.80 \
--context-length 131072 \
--max-running-requests 8 \
--cuda-graph-max-bs-decode 8 \
--min-free-slots-delay 1 \
--sampling-backend flashinfer \
--reasoning-parser qwen3 \
--tool-call-parser qwen3_coder \
--trust-remote-code
```
## Acknowledgements
We thank [DFlash](https://github.com/z-lab/dflash), [SpecForge](https://github.com/sgl-project/SpecForge), [SGLang](https://github.com/sgl-project/sglang) and [FlashInfer](https://github.com/flashinfer-ai/flashinfer). Dynamic verification references [DSpark](https://arxiv.org/abs/2607.05147).
## Citation
```bibtex
@article{zhang2026dspine,
title = {Draft in Parallel, Condition Through Depth: Adjacent Causal Injection for Speculative Decoding},
author = {Zhang, Haohui and Chen, Keyu and Sun, Haocheng and Gu, Weibo and Qiao, Ruizhi and Sun, Xing and Jiang, Bo},
journal = {arXiv preprint arXiv:2609.36173},
year = {2026}
}
```
## License
[Apache 2.0](LICENSE)