Arao AI
Back to search results

gemma-4-E2B-it · IQ2_M

Start a discussion
NVIDIA DGX Spark128GBvLLM v0.26L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-E2B-it (IQ2_M) on NVIDIA DGX Spark with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

380.91 tok/s

Decode

Decode speed

2,066.12 tok/s

Prefill

Prefill speed

0.4 s

TTFT

Time to first token

4.28 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. vLLM · IQ2_M (current)Decode 380.91 · Prefill 2066.12 ·
  2. vLLM · IQ3_XS Decode 360.01 · Prefill 1436.84 ·
  3. Strata · Q3_K_L Decode 389.31 · Prefill 1363.67 ·
  4. Strata · Q4_K_M Decode 326.71 · Prefill 1358.52 ·

Core figures

Decode (avg)
380.91 tok/s
Prefill (avg)
2,066.12 tok/s
TTFT (avg)
0.4 s
VRAM (avg)
4.28 GB
MTP acceptance rate
—
TTFB
3.05 GB
Power draw
167.3 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-E2B-it
Quantization
IQ2_M
Framework
vLLM
Version
v0.26
Context length
8192 tokens
Batch size
16
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)
4.28 GB
OS
Ubuntu 24.04
Driver
565.57
CUDA
12.4
Power draw
167.3 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

    v0.26 · Ubuntu 24.04 · CUDA 12.4 · 8192 ctx

    380.91 tok/s

    Decode

    2,066.12 tok/s

    Prefill

    0.4 s

    TTFT

    4.28 GB

    VRAM

    —

    MTP

    167.3 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=54 running model_id=66 at IQ2_M