Arao AI
Back to search results

Qwen3.8-9B · IQ3_S

Start a discussion
Apple M5 Pro64GBvLLM v0.31L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.8-9B (IQ3_S) on Apple M5 Pro with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

158.08 tok/s

Decode

Decode speed

657.93 tok/s

Prefill

Prefill speed

2.07 s

TTFT

Time to first token

14.57 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. vLLM · IQ3_S (current)Decode 158.08 · Prefill 657.93 ·
  2. vLLM · Q2_K Decode 169.33 · Prefill 690.81 ·
  3. Ollama · IQ3_S Decode 136.51 · Prefill 597.98 ·
  4. Ollama · Q3_K_S Decode 144.08 · Prefill 671.77 ·

Core figures

Decode (avg)
158.08 tok/s
Prefill (avg)
657.93 tok/s
TTFT (avg)
2.07 s
VRAM (avg)
14.57 GB
MTP acceptance rate
—
TTFB
1.26 GB
Power draw
60.9 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.8-9B
Quantization
IQ3_S
Framework
vLLM
Version
v0.31
Context length
32768 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
Apple M5 Pro
Nominal VRAM
64 GB
Measured VRAM (avg)
14.57 GB
OS
macOS 15.0
Driver
Apple M5 Pro
CUDA
N/A
Power draw
60.9 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-13

    v0.31 · macOS 15.0 · CUDA N/A · 32768 ctx

    158.08 tok/s

    Decode

    657.93 tok/s

    Prefill

    2.07 s

    TTFT

    14.57 GB

    VRAM

    —

    MTP

    60.9 W

    Power draw

    L2 batch seed 2026-10-13: vLLM on hw_id=27 running model_id=52 at IQ3_S