--- base_model: - z-lab/Qwen3-8B-DFlash-b16 license: mit pipeline_tag: text-generation library_name: transformers --- # Qwen3-8B-Ins-Draft-OPD This repository contains **Qwen3-8B-Ins-Draft-OPD**, a draft model for speculative decoding. The model is post-trained from [`z-lab/Qwen3-8B-DFlash-b16`](https://huggingface.co/z-lab/Qwen3-8B-DFlash-b16). It keeps the overall architecture and inference interface consistent with the original DFlash draft model, while further adapting the draft model through the Draft-OPD post-training method. - **Paper:** [Draft-OPD: On-Policy Distillation for Speculative Draft Models](https://huggingface.co/papers/2605.29343) - **Project Page:** [https://www.haodilei.top/draft-opd/](https://www.haodilei.top/draft-opd/) - **Repository:** [https://github.com/bingyang-lei/Draft-OPD](https://github.com/bingyang-lei/Draft-OPD) ## Model Details - **Target Model:** [`Qwen3-8B(enable_thinking=False)`](https://huggingface.co/Qwen/Qwen3-8B) - **Model type:** Draft model for speculative decoding - **Architecture:** Same as the original DFlash draft model - **Post-training method:** Draft-OPD ## Performance and Training Method Draft-OPD (On-Policy Distillation) trains speculative draft models with on-policy target feedback. Instead of only learning from fixed target-generated trajectories, the drafter is supervised on draft-induced states exposed during speculative verification, including the positions where draft proposals are rejected. Experiments show that Draft-OPD achieves over 5x lossless acceleration for thinking models across diverse tasks, improving over previous methods like EAGLE-3 and DFlash. For detailed training procedures, evaluation settings, and performance results, please refer to the paper. ## Citation If you find our work useful, please consider citing our paper: ```bibtex @misc{lei2026draftopdonpolicydistillationspeculative, title={Draft-OPD: On-Policy Distillation for Speculative Draft Models}, author={Haodi Lei and Yafy Li and Haoran Zhang and Shunkai Zhang and Qianjia Cheng and Xiaoye Qu and Ganqu Cui and Bowen Zhou and Ning Ding and Yun Luo and Yu Cheng}, year={2026}, eprint={2605.29343}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2605.29343}, } ```
