Arao AI
Back to search results

gemma-4-E4B-it · Q3_K_S

Start a discussion
NVIDIA H100 80GB HBM380GBvLLM v0.39L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-E4B-it (Q3_K_S) on NVIDIA H100 80GB HBM3 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

343.31 tok/s

Decode

Decode speed

1,851.05 tok/s

Prefill

Prefill speed

0.94 s

TTFT

Time to first token

13.02 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

12 days ago

Last verified

Performance

  1. llama.cpp · IQ3_XS Decode 297.43 · Prefill 1240.57 ·
  2. vLLM · Q3_K_S (current)Decode 343.31 · Prefill 1851.05 ·

Core figures

Decode (avg)
343.31 tok/s
Prefill (avg)
1,851.05 tok/s
TTFT (avg)
0.94 s
VRAM (avg)
13.02 GB
MTP acceptance rate
—
TTFB
2.75 GB
Power draw
674.6 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-E4B-it
Quantization
Q3_K_S
Framework
vLLM
Version
v0.39
Context length
16384 tokens
Batch size
32
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 H100 80GB HBM3
Nominal VRAM
80 GB
Measured VRAM (avg)
13.02 GB
OS
Ubuntu 24.04
Driver
572.83
CUDA
12.8
Power draw
674.6 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-24

    v0.39 · Ubuntu 24.04 · CUDA 12.8 · 16384 ctx

    343.31 tok/s

    Decode

    1,851.05 tok/s

    Prefill

    0.94 s

    TTFT

    13.02 GB

    VRAM

    —

    MTP

    674.6 W

    Power draw

    L2 batch seed 2026-10-24: vLLM on hw_id=55 running model_id=67 at Q3_K_S