Arao AI
Back to search results

Qwen3-4B-Thinking-2507 · IQ4_XS

Start a discussion
AMD Strix Point (Radeon 890M)llama.cpp v0.24L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B-Thinking-2507 (IQ4_XS) on AMD Strix Point (Radeon 890M) with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

75.65 tok/s

Decode

Decode speed

261.83 tok/s

Prefill

Prefill speed

2.44 s

TTFT

Time to first token

8.32 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 · IQ4_XS (current)Decode 75.65 · Prefill 261.83 ·
  2. vLLM · Q5_K_S Decode 60.14 · Prefill 291.94 ·

Core figures

Decode (avg)
75.65 tok/s
Prefill (avg)
261.83 tok/s
TTFT (avg)
2.44 s
VRAM (avg)
8.32 GB
MTP acceptance rate
—
TTFB
0.61 GB
Power draw
104.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B-Thinking-2507
Quantization
IQ4_XS
Framework
llama.cpp
Version
v0.24
Context length
32768 tokens
Batch size
512
GPU layers
99
Flash Attention
Off

Hardware

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

GPU
AMD Strix Point (Radeon 890M)
Measured VRAM (avg)
8.32 GB
OS
Ubuntu 22.04
Driver
565.57
CUDA
N/A
Power draw
104.4 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

    v0.24 · Ubuntu 22.04 · CUDA N/A · 32768 ctx

    75.65 tok/s

    Decode

    261.83 tok/s

    Prefill

    2.44 s

    TTFT

    8.32 GB

    VRAM

    —

    MTP

    104.4 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=89 running model_id=57 at IQ4_XS