Arao AI
Back to search results

gemma-4-E4B-it · FP16

Start a discussion
NVIDIA DGX Spark128GBvLLM v0.22L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-E4B-it (FP16) on NVIDIA DGX Spark with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

107.3 tok/s

Decode

Decode speed

444.7 tok/s

Prefill

Prefill speed

1.39 s

TTFT

Time to first token

17.64 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. llama.cpp · Q2_K Decode 258.21 · Prefill 920.61 ·
  2. vLLM · FP16 (current)Decode 107.3 · Prefill 444.7 ·

Core figures

Decode (avg)
107.3 tok/s
Prefill (avg)
444.7 tok/s
TTFT (avg)
1.39 s
VRAM (avg)
17.64 GB
MTP acceptance rate
—
TTFB
0.86 GB
Power draw
155.9 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-E4B-it
Quantization
FP16
Framework
vLLM
Version
v0.22
Context length
8192 tokens
Batch size
8
GPU layers
999
Flash Attention
On

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
NVIDIA DGX Spark
Nominal VRAM
128 GB
Measured VRAM (avg)
17.64 GB
OS
Windows 11 23H2
Driver
555.85
CUDA
12.4
Power draw
155.9 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-16

    v0.22 · Windows 11 23H2 · CUDA 12.4 · 8192 ctx

    107.3 tok/s

    Decode

    444.7 tok/s

    Prefill

    1.39 s

    TTFT

    17.64 GB

    VRAM

    —

    MTP

    155.9 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=54 running model_id=67 at FP16