{"version":1,"type":"story","url":"https://digestai.news/story/inco-ai-releases-dflash-2-draft-model-for-qwen3-8-flash-next","json":"https://digestai.news/story/inco-ai-releases-dflash-2-draft-model-for-qwen3-8-flash-next.json","markdown":"https://digestai.news/story/inco-ai-releases-dflash-2-draft-model-for-qwen3-8-flash-next.md","slug":"inco-ai-releases-dflash-2-draft-model-for-qwen3-8-flash-next","headline":"Inco AI releases DFlash 2 draft model for Qwen3.8-Flash-Next","summary":"DFlash 2 draft model is released as a repository that runs inside a speculative decoding server and drafts tokens for the target model Qwen3.8-Flash-Next. It predicts a whole block of tokens in a single pass, keeps the top candidates at every position, and uses a lightweight selector to trace one coherent path. Two‑tap dynamic convolutions in the backbone keep the draft from decaying toward the end of the block. Decoding is lossless: greedy output matches the target model exactly and sampling preserves its distribution. The model is served with SGLang and requires a small target‑feature capture patch that has been validated at TP1.\n\nThe target is the official FP8 checkpoint Qwen/Qwen3.8-Flash-Next-FP8. Runtime uses SGLang on one NVIDIA GB300 (TP1). Speculation block size is 4, giving three draft tokens per verification step, limited by the current QSA implementation. Sampling follows the officially recommended parameters: temperature 1.0, top‑p 0.95, top‑k 20, with a maximum of 4096 new tokens. Benchmarks include 128 samples at concurrency 1 and 1,024 samples at concurrency 8 and 32, comparing autoregressive decoding, Flash‑Next’s built‑in MTP, and DFlash 2, with token‑per‑second speedups reported.","keyPoints":["dflash 2 drafts up to 3 tokens per verification step with block size 4","sampling uses temperature 1.0, top‑p 0.95, top‑k 20 and max 4096 new tokens","benchmarks run 128 samples at concurrency 1 and 1,024 at concurrency 8 and 32 on an nvidia gb300"],"whyItMatters":"Block‑diffusion drafting can cut inference latency for large models like qwen3.8‑flash‑next, making real‑time applications more feasible.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Inco AI","SGLang","NVIDIA"],"models":["Qwen3.8-Flash-Next","Qwen3.8-Flash-Next-FP8","DFlash 2"],"people":[]},"firstPublishedAt":"2026-10-19T08:55:02Z","updatedAt":"2026-10-19T08:55:02Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"huggingface.co","title":"incoai/Qwen3.8-Flash-Next-DFlash2","url":"https://huggingface.co/incoai/Qwen3.8-Flash-Next-DFlash2","publishedAt":"2026-10-19T08:55:02Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[{"site":"Reddit","url":"https://www.reddit.com/r/LocalLLaMA/comments/1x9rq3r/incoaiqwen38flashnextdflash2/","points":null}],"thread":null,"cite":{"text":"Digest AI, \"Inco AI releases DFlash 2 draft model for Qwen3.8-Flash-Next\", 19 October 2026, https://digestai.news/story/inco-ai-releases-dflash-2-draft-model-for-qwen3-8-flash-next","publisher":"Digest AI","title":"Inco AI releases DFlash 2 draft model for Qwen3.8-Flash-Next","datePublished":"2026-10-19T08:55:02Z","url":"https://digestai.news/story/inco-ai-releases-dflash-2-draft-model-for-qwen3-8-flash-next"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}