FR
Réserver un appel

Local AI · Recipe · 2× DGX Spark

Qwen3.8-Flash-Next MLX 4-bit with TensorFold on 2× DGX Spark

Qwen3.8-Flash-Next MLX 4-bit with TensorFold on 2× DGX Spark: 69.6 tok/s according to github.com (dataset as of Sep 29, 2026).

by sfxnz

69.6tok/sEveryday

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

Source: github.com
Intelligence (original model) · Artificial Analysis40with thinking
Instructions in the README Repo

Engine

Engine
TensorFold v0.6.2 (56e2e3e), TP=2, Image tf-qwen38-flashnext:0.6.2 (nvcr.io/nvidia/pytorch:26.07-py3)
Quantization
MLX affine 4-bit (group 32) auf allen Linears inkl. n-gram-Tabellen und MTP-Head, KV bf16
Model family
Qwen3.8-Flash-Next
Context
262,144
Parameters
~180B MoE (512 Experten, top-10)
Creator
sfxnz
GitHub stars
1
Repo updated
Oct 2, 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)69.6 tok/sBench_decode.py (frozen ruler), streamed greedy, thinking off, max_tokens 200, 3-run-Median; MTP bis 15 Drafts, conf 0.70; TTFT p50 0.05 s Source 
  • Peak (decode)250 tok/sStructured c=1 (frozen ruler); opt-in PROFILE=concurrent (PR-#141-Patch): 519.6 tok/s aggregiert bei 8 Usern (Code, greedy) Source 

Prefill by context

  • 32K2,816 tok/sKalt, 32.921 Prompt-Tokens, 2.816–2.817 tok/s; 8k: 2.811–2.830, 131k: 2.377–2.378 Source 

Time to first token

Shorter is better.

  • Prompt not stated50 msProse c=1, TTFT p50 0.05 s Source 

What you need

Hardware
2 × NVIDIA DGX Spark (GB10)
Engine
TensorFold v0.6.2 (56e2e3e), TP=2, Image tf-qwen38-flashnext:0.6.2 (nvcr.io/nvidia/pytorch:26.07-py3)
Context
262,144 tokens

Sources

Related recipes

2× DGX Spark

Qwen3.8-Flash-Next 125B INT4-AutoRound with TensorFold on 2× DGX Spark (1M context)

by MiaAI-Lab

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 8, 2026

Details
2× DGX Spark

Qwen3.8-Flash-Next INT4-AutoRound (v5.2) with vLLM on 2× DGX Spark

by myllmbox

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 6, 2026

Details
2× DGX Spark

Qwen3.8-Flash-Next 125B NVFP4 with TensorFold on 2× DGX Spark (1M context)

by MiaAI-Lab

67.4tok/sEveryday

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

Intelligence40with thinking
  • Custom kernel required
  • License unclear

NVFP4 (nvidia, Experten) + FP8 n-gram/MTP, 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-DraftsWeights Repo Updated Oct 8, 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.

See the AI workshopBack to overview