Arao AI
Back to search results

Qwen3-8B · Q4_K_M

Start a discussion
AMD Strix Point (Radeon 890M)vLLM v0.42L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-8B (Q4_K_M) on AMD Strix Point (Radeon 890M) with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

56.8 tok/s

Decode

Decode speed

217.7 tok/s

Prefill

Prefill speed

1.76 s

TTFT

Time to first token

15.2 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 · Q4_K_M (current)Decode 56.8 · Prefill 217.7 ·
  2. Strata · IQ1_S Decode 121.73 · Prefill 484.23 ·
  3. Strata · Q5_K_S Decode 66.77 · Prefill 312.73 ·

Core figures

Decode (avg)
56.8 tok/s
Prefill (avg)
217.7 tok/s
TTFT (avg)
1.76 s
VRAM (avg)
15.2 GB
MTP acceptance rate
—
TTFB
0.45 GB
Power draw
83.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-8B
Quantization
Q4_K_M
Framework
vLLM
Version
v0.42
Context length
2048 tokens
Batch size
32
GPU layers
999
Flash Attention
On

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)
15.2 GB
OS
Windows 11 23H2
Driver
560.94
CUDA
N/A
Power draw
83.5 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

    v0.42 · Windows 11 23H2 · CUDA N/A · 2048 ctx

    56.8 tok/s

    Decode

    217.7 tok/s

    Prefill

    1.76 s

    TTFT

    15.2 GB

    VRAM

    —

    MTP

    83.5 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=89 running model_id=53 at Q4_K_M