DFlash
Block diffusion draft models for faster LLM inference
The Panel's Take
DFlash applies block diffusion models as draft generators for speculative decoding of autoregressive LLMs. Instead of predicting one token at a time, a small diffusion-based draft model generates multiple candidate tokens simultaneously — then the target LLM verifies them in parallel. The result is meaningfully faster inference with no loss in output quality. The library is compatible with all major inference serving frameworks: vLLM, SGLang, Hugging Face Transformers, and MLX (for Apple Silicon). It ships with 15+ pretrained draft models on HuggingFace covering popular base models. The underlying research (arXiv:2602.06036) has been validated with support from NVIDIA and Modal Labs, suggesting production viability. The repo was trending on GitHub with 280+ new stars. Speculative decoding has been one of the most practical LLM speed-up techniques of the past two years, but finding good draft models has always been painful. DFlash's diffusion approach sidesteps the need for a carefully size-matched autoregressive draft model, potentially making speculative decoding accessible to a wider range of deployed models.
Share this verdict
DFlash verdict: SHIP 🚀 3 ships · 1 skip from the expert panel Full review: shiporskip.io/tool/dflash-block-diffusion-speculative-decoding-vllm-sglang-2026
Weekly AI Tool Verdicts
Get the next verdict in your inbox
7 critics review a new AI tool every day. Weekly digest — free.
Compare DFlash with Others
Embed this verdict
Tool makers can add a live ShipOrSkip badge to their site. Badge loads track impressions; clicks route back to this review.
<a href="https://shiporskip.io/api/badge-click/dflash-block-diffusion-speculative-decoding-vllm-sglang-2026" target="_blank" rel="noopener"><img src="https://shiporskip.io/api/badge/dflash-block-diffusion-speculative-decoding-vllm-sglang-2026" alt="DFlash Ship verdict on ShipOrSkip" width="360" height="90" /></a>[](https://shiporskip.io/api/badge-click/dflash-block-diffusion-speculative-decoding-vllm-sglang-2026)<iframe src="https://shiporskip.io/embed/dflash-block-diffusion-speculative-decoding-vllm-sglang-2026" title="DFlash ShipOrSkip verdict" width="360" height="260" style="border:0;border-radius:16px;max-width:100%;" loading="lazy"></iframe>The reviews
“vLLM and SGLang integration out of the box means I can drop this into an existing serving stack without a rewrite. The 15+ pretrained draft models remove the biggest friction point of speculative decoding setups. If the benchmarks hold in production, this is an easy win for latency-sensitive deployments.”
“Speculative decoding speedups are notoriously workload-dependent — they shine on long completions and suffer on short ones. Diffusion-based drafts add another variable: acceptance rates depend on how well the draft distribution matches your target model's. Real-world numbers on diverse prompts are what I need before calling this a universal win.”
“Inference efficiency compounds over time — every latency improvement at the serving layer makes more agentic applications economically viable. DFlash's approach of using diffusion models as universal draft generators could become the default speculative decoding strategy once the acceptance rates mature.”
“Faster inference means snappier AI tools for everyone. I don't care about the underlying math — I care that my AI writing assistant responds in under a second. If DFlash helps the infra teams get there, I'm all for it shipping.”