Arao AI
Back to search results

Qwen3-8B · IQ1_S

Start a discussion
AMD Strix Halo (Radeon 8060S)vLLM v0.32L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-8B (IQ1_S) on AMD Strix Halo (Radeon 8060S) with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

127.84 tok/s

Decode

Decode speed

540.46 tok/s

Prefill

Prefill speed

1.79 s

TTFT

Time to first token

9.16 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 · IQ1_S (current)Decode 127.84 · Prefill 540.46 ·
  2. vLLM · Q4_K_S Decode 94.74 · Prefill 425.68 ·
  3. Ollama · Q5_K_M Decode 64.59 · Prefill 334.15 ·
  4. Ollama · Q8_0 Decode 48.28 · Prefill 205.87 ·

Core figures

Decode (avg)
127.84 tok/s
Prefill (avg)
540.46 tok/s
TTFT (avg)
1.79 s
VRAM (avg)
9.16 GB
MTP acceptance rate
—
TTFB
1.02 GB
Power draw
107.3 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-8B
Quantization
IQ1_S
Framework
vLLM
Version
v0.32
Context length
8192 tokens
Batch size
8
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
AMD Strix Halo (Radeon 8060S)
Measured VRAM (avg)
9.16 GB
OS
macOS 15.0
Driver
560.94
CUDA
N/A
Power draw
107.3 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

    v0.32 · macOS 15.0 · CUDA N/A · 8192 ctx

    127.84 tok/s

    Decode

    540.46 tok/s

    Prefill

    1.79 s

    TTFT

    9.16 GB

    VRAM

    —

    MTP

    107.3 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=87 running model_id=53 at IQ1_S