Arao AI
Back to search results

Qwen3-4B-Thinking-2507 · Q3_K_S

Start a discussion
Qualcomm Snapdragon 7 Gen 4llama.cpp v0.53L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B-Thinking-2507 (Q3_K_S) on Qualcomm Snapdragon 7 Gen 4 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

53.5 tok/s

Decode

Decode speed

273.17 tok/s

Prefill

Prefill speed

1.26 s

TTFT

Time to first token

7.56 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 · Q3_K_S (current)Decode 53.5 · Prefill 273.17 ·

Core figures

Decode (avg)
53.5 tok/s
Prefill (avg)
273.17 tok/s
TTFT (avg)
1.26 s
VRAM (avg)
7.56 GB
MTP acceptance rate
—
TTFB
0.43 GB
Power draw
15.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B-Thinking-2507
Quantization
Q3_K_S
Framework
llama.cpp
Version
v0.53
Context length
2048 tokens
Batch size
256
GPU layers
33
Flash Attention
On

Hardware

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

GPU
Qualcomm Snapdragon 7 Gen 4
Measured VRAM (avg)
7.56 GB
OS
Windows 11 23H2
Driver
Qualcomm Adreno 7.0
CUDA
N/A
Power draw
15.4 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

    v0.53 · Windows 11 23H2 · CUDA N/A · 2048 ctx

    53.5 tok/s

    Decode

    273.17 tok/s

    Prefill

    1.26 s

    TTFT

    7.56 GB

    VRAM

    —

    MTP

    15.4 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=105 running model_id=57 at Q3_K_S