specific method · filed under inference & serving
DFlash
A speculative-decoding drafter using a lightweight block-diffusion model to produce a draft block in one forward pass.
Also called Block-6 DFlash speculative decoding, block-6 DFlash for speculative decoding, DFlash speculative decoding, DFlash-style speculative decoding drafter, DFlash-style, 5-layer SWA MTP drafter (DFlash-style).
- sources
- 6
- models
- 4
- labs adopt it
- 3
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
Documented in
Further reading
Picked by hand, not extracted: where to read more, not evidence for anything on this page.
- DFlash: Block Diffusion for Flash Speculative Decoding paper arxiv.orgintroduces the DFlash draft model
Evidence
8 spans quoted from the sources, strongest treatment first.
5-layer SWA MTP drafter (DFlash-style). Predicts 7 subsequent tokens per forward pass for parallel verification.
Speculative decoding in MiMo-V2.6 uses a multi-token prediction (MTP) module following the block diffusion design of DFlash
it predicts 7 subsequent tokens in a single forward pass for parallel verification by the backbone
5-layer SWA MTP drafter (DFlash-style). Predicts 7 subsequent tokens per forward pass for parallel verification.
RL rollouts use block-6 DFlash for speculative decoding by default, replacing the multi-token prediction (MTP-3) configuration inherited from SFT.
NVIDIA is also releasing a DFlash draft model, which can be measured against the others and may perform best for your workloads.
DFlash: A speculative-decoding drafter that uses a lightweight block-diffusion model to generate an entire draft block in one forward pass.
Speculative decoding: a trained DFlash draft model is available for lower-latency serving
Filed alongside
Other methods under inference & serving :: decoding strategy.