[image] DFlash 2: Keep Drafting Parallel DFlash 2 is the successor to our widely deployed parallel drafter: close to 3× the speed of autoregressive decoding, with the same output. Drafters for Qwen3.8-27B and Meta's Muse Glimmer are out today. Has anyone trie…

just trying this one: GitHub - Weschera/Qwen3.8-27B-NVFP4-DFlash2-DGX-Spark: Reproducible Qwen3.8-27B NVFP4 + DFlash2 recipe and qualification evidence for one NVIDIA DGX Spark · GitHub

And this one too, I have been using Mia-Lab’s regular MTP branch, works well. [image] GitHub - MiaAI-Lab/Qwen3.8-27B-SGLang-DGX-Spark: Qwen3.8 27B on SGLang for DGX Spark Qwen3.8 27B on SGLang for DGX Spark You will need to change the reasoning effort…

Finally something who gives a real difference, now it’s a little bit more comfortable with Dflash2 : | model | test | t/s | peak t/s | ttfr (ms) | est_ppt (ms) | e2e_ttft (ms) | |:--------|---------------:|-----------------:|--------------:|--------------…

I tried applying the path using eugr vllm but it’s complaining about “hf_overrides must be a dict for get_quant_config to get the quantization config from it.” I’ll try rebuilding the container again from scratch and applying the pr again, maybe I missed something

How does DFlash2 performance scale compared to MTP? At high context and/or high concurrency, are you guys seeing comparable performance for both PP and TG?

Currently waiting for mainline SGLang release to have DFlash2 support incorporated, with qwen38 support as well. There are some workarounds where you will have to build your own docker image. I would rather wait for main of SGLang to catch up and use the official docker image. Currently using MTP, …

After running it through my benchmarking tasks (simulated after what I do at work) I get a jump from dspark ~25 to dflash2 ~38tps (average weighted by spans of sglang actually giving me tokens, stuff like waiting for tool calls excluded). It’s finally usable. MTP was slightly worse than dspark.

I decided to evaluate if 27B FP8 can be used as a main model on 2 sparks - I used it successfully on Metal but given the context and parameter limitation - not sure for 2 sparks. Tested Mia’s SGLang config in all possible variation. Final result is real: ╭──────────────────────────────────────────…

And another one for good measure - now with YaRN enabled and max length set to 500k - quality expected to take a hit but still ╭───────────────────────────────────────────────────────────────────────── 🏆 Benchmark Complete ─────────────────────────────────────────────────────────────────────────╮ │…

Qwen 3.8 27B + DFlash2

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10

danilo.luvizotto August 19, 2026, 6:22pm 4

yes, I’m getting 40-42 tok/s

Topic		Replies	Views
Qwen3.8-27B (NVFP4) on Single/Dual DGX Spark — SGLang + DFlash2, fully OpenAI-compatible DGX Spark / GB10	15	5848	September 6, 2026
Qwen3.8-27B on dual Sparks DGX Spark / GB10 agentic-ai	17	6534	August 22, 2026
Qwen3.8-27B at 34–38 tok/s on DGX Spark — open-source one-command setup (SGLang + NVFP4 + DSpark) DGX Spark / GB10 Projects llm , llama , agentic-ai , deepseek	106	19213	September 16, 2026
Qwen3.8-27B benchmarking on one DGX Spark: DFlash2 beat vLLM+MTP, and greedy beat the "thinking" sampler DGX Spark / GB10 Projects benchmarks , llm , agentic-ai	24	5465	September 7, 2026
Comprehensive Qwen3.8-27B Study on DGX Sparks: Quantization, Speculative Decoding, and TP/DP Scaling DGX Spark / GB10 Projects	10	4779	August 26, 2026
DFlash LLM for DGX Spark - too good to be true? DGX Spark / GB10	37	4398	April 17, 2026
Qwen3.8-Flash-Next on 1, 2 and 4 DGX Sparks with NVIDIA's official NVFP4 quant: 64 tok/s peak single stream DGX Spark / GB10	27	5072	September 10, 2026
Single DGX-Spark - Qwen 3.8-Flash-Next at ~43tok/sec in Coding DGX Spark / GB10	63	11922	September 14, 2026
1x Spark new workhorse: DFlash2 + SGLang is fairly fast for Qwen3.8 27B NVFP4 DGX Spark / GB10	3	1045	August 26, 2026
Qwen3.8-Flash-Next 180B, Single Solo DGX Spark With HashK-PLE NVFP4 DGX Spark / GB10 Projects	42	6375	September 5, 2026

Qwen 3.8 27B + DFlash2

Related topics