Local AI · Recipe · 1× DGX Spark

Qwen3.8-Flash-Next NVFP4 with vLLM on 1× DGX Spark

Qwen3.8-Flash-Next NVFP4 with vLLM on 1× DGX Spark: 76.2 tok/s according to github.com (dataset as of Nov 3, 2026).

by myllmbox

76.2tok/sPeak

1 request, code prompt (TypeScript), thinking off, v5.1 column (unchanged on the v5.2 image)

Source: github.com
Intelligence (original model) · Artificial Analysis40with thinking

Engine

Engine
vLLM 0.30 (RecoverSSM, dynamischer Draft-Depth bis 7, sampled Drafts mit Block-Verifikation, 16 Seats) — Kit v5.2; hibrid48 bleibt als Alternative (~20 % langsamer als INT4-Default)
Quantization
hibrid48: kalibrierter Body mit NVFP4-Output-Head (0.33 GiB), bf16-KV (27G-Pin), NVFP4-n-gram-Tabelle, Marlin MoE
Model family
Qwen3.8-Flash-Next
Context
262,144
Creator
myllmbox
GitHub stars
169
Repo updated
Oct 30, 2026

Measurements

Every sourced value of this recipe, each with its condition and source. Bars relative to the largest value in the group.

Decode · 1 request

  • Everyday (decode)55.6 tok/sMixed (code + explanation), c=1, Thinking off (v5.1-Spalte) Source 
  • Peak (decode)76.2 tok/sCode (TypeScript), c=1, Thinking off (v5.1-Spalte); Thinking-on (12-Run-Tabelle): 50,9 / 71,7 / 90,8 (thinking/code/peak); Peak @16 Streams (Thinking off, mixed): 335 aggregiert Source 

What you need

Hardware
1 × NVIDIA DGX Spark (GB10)
Engine
vLLM 0.30 (RecoverSSM, dynamischer Draft-Depth bis 7, sampled Drafts mit Block-Verifikation, 16 Seats) — Kit v5.2; hibrid48 bleibt als Alternative (~20 % langsamer als INT4-Default)
Context
262,144 tokens

Related recipes

2× DGX Spark
89.3tok/sEveryday

1 request, prose prompt, with MTP (per recipe) Source 

Intelligenceno independent value
  • Custom kernel required
  • License unclear

INT4-AutoRound (GPTQ int4, Top-5-Experten) + FP8 n-gram-Tabelle, FP8 KVTensorFold tensorfold-native (Zig, zig-flashnext db281878 + patches), Image ghcr.io/miaai-lab/qwen3.8-flash-dual-dgx-sparks-tensorfold:zig-db28187-f3e1300f4f62, TP=2 über CX7/RoCE, MTP + Copy-Drafts (QUANT=int4ar)Weights Repo Updated Oct 26, 2026

Details
2× DGX Spark
74tok/sEveryday

1 request, prose prompt (long story), thinking off, with MTP Source 

Intelligence40with thinking
  • Custom kernel required
  • License unclear

INT4-AutoRound (azampatti): int4-Experten, fp8-Seitenschichten; n-gram-Tabelle beim Laden fp8 → NVFP4, halb pro Box auf der GPU; KV-Pin 29G/Box (1.829.182 Tokens)vLLM 0.30 (Image myllmbox/qwen38-flash-next-cluster-vllm:v5.2; RecoverSSM, dynamischer Draft-Depth 3–7 in Full-CUDA-Graphs, sampled Drafts + Block-Verifikation, One-Shot-RoCE-All-Reduce, Dual-Rail, TP=2, 64 Seats) — Kit v5.2Weights Repo Updated Oct 29, 2026

Details

← Back to overview

Local AI in your company?

In a workshop we work out which models and which hardware fit your tasks, and build the first agent on your infrastructure.