Local AI · Recipe · 1× DGX Spark

Nemotron-3.5 30B NVFP4 with vLLM on 1× DGX Spark

Nemotron-3.5 30B NVFP4 with vLLM on 1× DGX Spark: 93 tok/s according to github.com (dataset as of Nov 3, 2026).

by sfxnz

93tok/sPeak

1 request, best value reported by source

Source: github.com
Intelligence (original model) · Artificial Analysis13with thinking

Engine

Engine
vLLM
Quantization
NVFP4 (modelopt_mixed), FP8-KV, Marlin MoE, DSpark-3
Model family
Nemotron-3.5
Context
262,144
Parameters
30B total / ~3B active (README); HF-API safetensors total 17,820,210,764
Creator
sfxnz
GitHub stars
1
Repo updated
Sep 4, 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

  • Peak (decode)93 tok/sL.A.I.L-Lab-Bench, Run id 20260812T131332Z_d0f7ae; Quelle '~93 tok/s' → approx; Messmethodik/Prompt im Repo nicht dokumentiert Source 

Prefill by context

  • Prompt not stated678 tok/sQuelle '~678 tok/s' → approx; L.A.I.L-Lab-Bench Source 

Time to first token

Shorter is better.

  • Prompt not stated246 msQuelle '~246 ms' (TTFT p50) → approx Source 

What you need

Hardware
1 × NVIDIA DGX Spark (GB10)
Engine
vLLM
Context
262,144 tokens

Related recipes

1× DGX Spark
71.8tok/sEveryday

1 request, prose prompt (sampled, thinking off), with LiLiCorr (per recipe) Source 

Intelligenceno independent value
  • ≤ 3 bit: quantization may cost quality
  • Custom kernel required
  • License unclear

24GB pack: IQ2_XXS experts (2.0625 bits), NVFP4 Mamba projections, Q8_0 attention/latent projections/lm_head; bf16 KVTensorFold tensorfold-native v1.1.6 (eea9dc9c) + recipe patches/ (Bolt lane backend, libbolt.so CUDA forward pass), LiLiCorr draftsWeights Repo Updated Nov 1, 2026

Details
1× DGX Spark
59.4tok/sEveryday

1 request, chat prompt (20 prompts, thinking off), no drafting (default profile) Source 

Intelligenceno independent value
  • Custom kernel required

EXL3 6.08 bpw (28,3 GB), FP16 KV; nur Textpfad (Vision-Tower wird von ExLlamaV3 nicht ausgeführt)ExLlamaV3 1.5.0 (Fork r0b0tlab/exllamav3-r0b0t, Branch minicpm-v4.7-dspark @ 5ddae7d12448), torch 2.14.0+cu130Weights Repo Updated Oct 14, 2026

Details
1× DGX Spark
55.6tok/sEveryday

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

Intelligence40with thinking
  • Custom kernel required
  • License unclear

NVIDIA FP4-QAD-Export (FP4-Local @b9ce18a6): Routed Experts NVFP4 (Gruppe 16), Attention/Shared Expert MXFP8, alle 512 Experten mit 10 pro Token; n-gram-Tabelle per pread von der SSD; FP8 KVTensorFold tensorfold-native (Zig) v1.1.4 (eab85828) + Patches 0001–0017, Image tensorfold-qwen38:v1.1.4-eab8582 (GHCR v1.1.4-eab8582-4baf16137ecd) auf nvcr.io/nvidia/pytorch:26.07-py3, MTP-DraftsWeights Repo Updated Oct 30, 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.