DFlash Speculative Decoding Drafts Whole Token Blocks in Parallel for Up to 15x Higher Throughput on NVIDIA Blackwell
15x throughput. That's what DFlash achieves on NVIDIA Blackwell by replacing autoregressive drafting with parallel block diffusion—and it's shipping to SGLang, vLLM, and TensorRT-LLM today.
Why it matters
DFlash introduces a fundamental shift in speculative decoding architecture, moving from token-by-token drafting to parallel block generation. This unlocks measurable throughput gains on production inference infrastructure (Blackwell), directly impacting deployment economics for inference-heavy applications.
The key facts
7 to know6.08x lossless speedup on Qwen3-8B (paper results)
Up to 15x throughput on NVIDIA Blackwell (at fixed interactivity)
Block diffusion model replaces autoregressive drafting
KV injection conditions on target hidden features
20 checkpoints shipped
Integrations: SGLang, vLLM, TensorRT-LLM
UC San Diego research
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: UC San Diego's DFlash replaces autoregressive drafting with a lightweight block diffusion model for speculative decoding. It drafts whole token blocks in a single forward pass and conditions on target hidden features through KV injection. The paper reports up to 6.08x lossless speedup on Qwen3-8B,…