Arao AI
Back to search results

Qwen3.5-4B · Q8_0

Start a discussion
AMD Strix Halo (Radeon 8060S)vLLM v0.49L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-4B (Q8_0) on AMD Strix Halo (Radeon 8060S) with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

72.19 tok/s

Decode

Decode speed

307.07 tok/s

Prefill

Prefill speed

1.1 s

TTFT

Time to first token

11.76 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 · Q8_0 (current)Decode 72.19 · Prefill 307.07 ·
  2. Ollama · IQ3_XS Decode 123.41 · Prefill 403.1 ·

Core figures

Decode (avg)
72.19 tok/s
Prefill (avg)
307.07 tok/s
TTFT (avg)
1.1 s
VRAM (avg)
11.76 GB
MTP acceptance rate
—
TTFB
0.58 GB
Power draw
94.6 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-4B
Quantization
Q8_0
Framework
vLLM
Version
v0.49
Context length
2048 tokens
Batch size
8
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 Halo (Radeon 8060S)
Measured VRAM (avg)
11.76 GB
OS
Ubuntu 22.04
Driver
565.57
CUDA
N/A
Power draw
94.6 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

    v0.49 · Ubuntu 22.04 · CUDA N/A · 2048 ctx

    72.19 tok/s

    Decode

    307.07 tok/s

    Prefill

    1.1 s

    TTFT

    11.76 GB

    VRAM

    —

    MTP

    94.6 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=87 running model_id=48 at Q8_0