Arao AI
Back to search results

gemma-4-E4B-it · IQ2_XS

Start a discussion
NVIDIA H2097GBllama.cpp v0.35L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-E4B-it (IQ2_XS) on NVIDIA H20 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

95.17 tok/s

Decode

Decode speed

436.06 tok/s

Prefill

Prefill speed

1.22 s

TTFT

Time to first token

10.16 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

11 days ago

Last verified

Performance

  1. llama.cpp · IQ2_XS (current)Decode 95.17 · Prefill 436.06 ·
  2. vLLM · Q3_K_M Decode 78.41 · Prefill 259.38 ·

Core figures

Decode (avg)
95.17 tok/s
Prefill (avg)
436.06 tok/s
TTFT (avg)
1.22 s
VRAM (avg)
10.16 GB
MTP acceptance rate
—
TTFB
0.76 GB
Power draw
302 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-E4B-it
Quantization
IQ2_XS
Framework
llama.cpp
Version
v0.35
Context length
8192 tokens
Batch size
512
GPU layers
99
Flash Attention
On

Hardware

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

GPU
NVIDIA H20
Nominal VRAM
97 GB
Measured VRAM (avg)
10.16 GB
OS
Windows 11 23H2
Driver
572.83
CUDA
12.6
Power draw
302 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

    v0.35 · Windows 11 23H2 · CUDA 12.6 · 8192 ctx

    95.17 tok/s

    Decode

    436.06 tok/s

    Prefill

    1.22 s

    TTFT

    10.16 GB

    VRAM

    —

    MTP

    302 W

    Power draw

    L2 batch seed 2026-10-24: llama.cpp on hw_id=56 running model_id=67 at IQ2_XS