Arao AI
Back to search results

gemma-4-E4B-it · Q3_K_L

Start a discussion
Qualcomm Snapdragon X ElitevLLM v0.21L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-E4B-it (Q3_K_L) on Qualcomm Snapdragon X Elite with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

70.71 tok/s

Decode

Decode speed

337.65 tok/s

Prefill

Prefill speed

1.6 s

TTFT

Time to first token

8.15 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 · Q3_K_L (current)Decode 70.71 · Prefill 337.65 ·
  2. vLLM · Q4_K_M Decode 62.62 · Prefill 304.49 ·
  3. Ollama · Q3_K_L Decode 58.9 · Prefill 275.23 ·
  4. Ollama · Q8_0 Decode 35.6 · Prefill 173.17 ·

Core figures

Decode (avg)
70.71 tok/s
Prefill (avg)
337.65 tok/s
TTFT (avg)
1.6 s
VRAM (avg)
8.15 GB
MTP acceptance rate
—
TTFB
0.57 GB
Power draw
19.1 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-E4B-it
Quantization
Q3_K_L
Framework
vLLM
Version
v0.21
Context length
2048 tokens
Batch size
16
GPU layers
999
Flash Attention
Off

Hardware

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

GPU
Qualcomm Snapdragon X Elite
Measured VRAM (avg)
8.15 GB
OS
Windows 11 23H2
Driver
Qualcomm Adreno 7.0
CUDA
N/A
Power draw
19.1 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

    v0.21 · Windows 11 23H2 · CUDA N/A · 2048 ctx

    70.71 tok/s

    Decode

    337.65 tok/s

    Prefill

    1.6 s

    TTFT

    8.15 GB

    VRAM

    —

    MTP

    19.1 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=104 running model_id=67 at Q3_K_L