Arao AI
Back to search results

gemma-4-E4B-it · IQ2_M

Start a discussion
AMD Strix Halo (Radeon 8060S)llama.cpp v0.26L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-E4B-it (IQ2_M) on AMD Strix Halo (Radeon 8060S) with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

142.24 tok/s

Decode

Decode speed

506.53 tok/s

Prefill

Prefill speed

0.92 s

TTFT

Time to first token

6.47 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 · IQ2_M (current)Decode 142.24 · Prefill 506.53 ·
  2. Strata · Q8_0 Decode 92.35 · Prefill 312.7 ·

Core figures

Decode (avg)
142.24 tok/s
Prefill (avg)
506.53 tok/s
TTFT (avg)
0.92 s
VRAM (avg)
6.47 GB
MTP acceptance rate
—
TTFB
1.14 GB
Power draw
53.6 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-E4B-it
Quantization
IQ2_M
Framework
llama.cpp
Version
v0.26
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
AMD Strix Halo (Radeon 8060S)
Measured VRAM (avg)
6.47 GB
OS
Ubuntu 22.04
Driver
amdgpu 6.11
CUDA
N/A
Power draw
53.6 W

Sources & evidence

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

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

    v0.26 · Ubuntu 22.04 · CUDA N/A · 8192 ctx

    142.24 tok/s

    Decode

    506.53 tok/s

    Prefill

    0.92 s

    TTFT

    6.47 GB

    VRAM

    —

    MTP

    53.6 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=87 running model_id=67 at IQ2_M