Tencent DSpine’s five-layer speculative-decoding draft model for Qwen3.8-27B, using 16-token draft blocks and an 8K sliding window to propose tokens verified by the target model.
Official checkpoint ID: `tencent/Qwen3.8-27B-DSpine-b16`. It runs with the Qwen3.8-27B target through speculative decoding in SGLang; no standalone hosted API is listed.
Only published specifications are shown.
— Tencent’s official Hugging Face repository first uploaded a five-layer DSpine draft checkpoint for Qwen3.8-27B on 2026-11-07 under Apache 2.0. It is used for speculative decoding; it is not a new Qwen3.8 target-model version or a standalone hosted API. ↗
This is the first open-weight DSpine checkpoint targeting Qwen3.8-27B, adding a speculative-decoding path used alongside the base model. It does not replace Qwen3.8-27B and is not a standalone user-facing foundation model.
Tencent’s official Hugging Face repository first uploaded a five-layer DSpine draft checkpoint for Qwen3.8-27B on 2026-11-07 under Apache 2.0. It is used for speculative decoding; it is not a new Qwen3.8 target-model version or a standalone hosted API.
View sourceSelect a card to view the related model.
Copy this template, add the field, proposed value, and primary source, and share it with the archive maintainer.