Qwen3.8-Flash-Next-DFlash2
Qwen3.8-Flash-Next-DFlash2 is a block-diffusion draft model for speculative decoding with Qwen3.8-Flash-Next. It runs inside a speculative decoding server and is not a standalone language model.
What it does
Developers can use it with SGLang to accelerate Qwen3.8-Flash-Next inference while preserving the target model’s output behavior.
- Availability
- Repository contains the DFlash 2 draft model; quick start instructions use SGLang
- Pricing model
- Not established
- Deployment
- Serve with SGLang
- Privacy
- Not established
- Licence
- Apache-2.0
Capability
Modality
Access
Deployment
Pricing
Target user
Evidence
5 claims- DFlash 2 predicts a whole block of tokens in a single pass. [1]
a whole block of tokens in a single pass and keeps the top candidates at
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - Greedy output matches the target model exactly, and sampling preserves its distribution. [1]
matches the target model exactly, and sampling preserves its distribution.
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - The repository provides the draft model and quick-start instructions for serving it with SGLang. [1]
This repository contains the DFlash 2 draft model for
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - Flash-Next needs a target-feature capture patch included in the repository until support lands upstream. [1]
target-feature capture patch, [included in this repository](patches/qwen38-dflash2-capture.patch)
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - On one NVIDIA GB300 at concurrency 1, reported throughput speedups versus autoregressive decoding range from 1.63× to 1.82× across the listed tasks. [1]
Sources
1
1
Hugging FaceOwner-controlled
https://huggingface.co/incoai/Qwen3.8-Flash-Next-DFlash2
Retrieved 2026-10-27 01:06 UTC
Verified
Versions
3| Version | Effective | Change | Changed fields | State |
|---|---|---|---|---|
| 3 | 2026-10-22 00:04 UTC | evidence update | score | Current |
| 2 | 2026-10-20 16:08 UTC | material change | access, availability | Current |
| 1 | 2026-10-19 17:06 UTC | publication | None | Current |
In briefs
1- OpenGCUA for macOS leads 12 AI developments on 19 October 2026 Daily brief 2026-10-19
Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.