GLM-5.3-Flash gets DFlash2, a block-diffusion draft model for lossless speculative decoding
TheZachMueller · x · 2026-08-28
Inco AI released GLM-5.3-Flash-DFlash2 on Hugging Face, a speculative decoding draft model for GLM-5.3-Flash. It is not a standalone LLM — it runs inside a speculative decoding server and drafts tokens for the target model to verify.
How it works: DFlash 2 is a block-diffusion drafter that predicts a whole block of tokens in one pass, keeps top candidates at every position, and a lightweight selector traces one coherent path. Two-tap dynamic convolutions in the backbone keep drafts from decaying toward the end of the block. Decoding is lossless: greedy output matches the target exactly and sampling preserves its distribution.
Serving uses SGLang with --speculative-algorithm DFLASH; evaluation ran on 4x GB300 GPUs (TP4) with FP8 KV cache, speculation block size 8 (7 draft tokens per verification step), temperature 1.0 / top-p 0.95.
Related event: Inco AI Releases DFlash 2 Draft Model, Speeding Up GLM-5.3-Flash by 2.8x(3 posts)→
More from Infra
- a16z raises $1.1B Machine Age fund to invest in chips, robots, and AI hardware — KateClarkTweets · 2026-08-28
- US Commerce Dept to Block Chinese Access to Overseas NVIDIA Chips — McDonaghMatthew · 2026-08-28
- US data centers face 17-29 GW power load gap by 2026 — BenBajarin · 2026-08-28
- AI capex chases adoption, not leads it — YvesMulkers · 2026-08-28
- a16z on Why Infrastructure Is AI's New Bottleneck — a16z · 2026-08-28
- AI Chipmaker Sunrise Doubles Valuation After Raising 2 Billion Yuan — pstAsiatech · 2026-08-28