Radar / AI infrastructure / DSpine
DSpine: Tencent's parallel speculative-decoding drafter for Qwen3-8B
Tencent's Hunyuan team open-sourced DSpine — a parallel draft model for speculative decoding (arXiv 2609.36173) — with inference code and a bfloat16 Qwen3-8B drafter checkpoint under Apache 2.0 (repo created Oct 15, HF model Oct 16).
Why it matters
'Draft in parallel, condition through depth': each draft position receives its predecessor's predicted feature so the causal chain unfolds over network depth while all positions draft in parallel; tops DSpark, DFlash2 and Domino on Qwen3-8B acceptance length across GSM8K, HumanEval, Arena-Hard and more.
What you could build with it
A startup serving Qwen3-8B chatbots on cost-constrained GPUs could drop DSpine's drafter into SGLang via --speculative-algorithm DSPINE to raise tokens-per-second per GPU and cut serving cost without changing the base model's outputs.
Does it hold up?
Too early to judge — all acceptance-length numbers are from Tencent's own paper on Qwen3-8B; no third-party serving benchmarks or production deployments reported yet.
Built with DSpine
- tencent/dspinegithub · 4 ★ · Official Tencent repo with the drafter training/inference code and SGLang integration; no third-party integrations exist yet.
Learn more
First spotted on github: source.
More AI infrastructure
DeepGEMM: DeepSeek's open GPU BLAS kernel libraryDeepSeek's DeepGEMM, a clean and efficient open-source BLAS kernel library for NVIDIA GPUs (up to 1550 TFLOPS on…infra · JEV 0.77OpenViking — self-evolving context database for AI agentsByteDance Volcengine's open-source (AGPL-3.0) context database that unifies agent memory, knowledge RAG, and skills…infra · JEV 0.76Modular open-sources MAX inference server, Mojo stdlib and accelerator kernelsThe core of Modular's unified AI deployment stack is now open: the MAX inference server (OpenAI-compatible endpoints)…infra · JEV 0.76OmniRouteFree MIT AI gateway: one endpoint, 290+ providers and 500+ models with quota-aware auto-fallback.infra · JEV 0.75Context Mode: context-window optimization for AI coding agentsMCP server for AI coding agents that sandboxes verbose tool output and persists session memory, cutting context use by…infra · JEV 0.74StrataOpen-source inference engine that runs the 125B-parameter Qwen3.8-Flash-Next MoE on a single consumer GPU with 12GB+…infra · JEV 0.74
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.