Arao AI
Back to search results

Qwen3-1.7B · Q3_K_S

Start a discussion
Apple M2 ProvLLM v0.44L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-1.7B (Q3_K_S) on Apple M2 Pro with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

168.62 tok/s

Decode

Decode speed

623.7 tok/s

Prefill

Prefill speed

0.45 s

TTFT

Time to first token

4.42 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 · IQ3_XS Decode 176.88 · Prefill 629.62 ·
  2. llama.cpp · Q3_K_S Decode 168.44 · Prefill 689.11 ·
  3. vLLM · Q2_K Decode 177.75 · Prefill 635.73 ·
  4. vLLM · Q3_K_S (current)Decode 168.62 · Prefill 623.7 ·

Core figures

Decode (avg)
168.62 tok/s
Prefill (avg)
623.7 tok/s
TTFT (avg)
0.45 s
VRAM (avg)
4.42 GB
MTP acceptance rate
—
TTFB
1.35 GB
Power draw
16.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-1.7B
Quantization
Q3_K_S
Framework
vLLM
Version
v0.44
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 M2 Pro
Measured VRAM (avg)
4.42 GB
OS
macOS 15.0
Driver
Apple M2 Pro
CUDA
N/A
Power draw
16.7 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

    v0.44 · macOS 15.0 · CUDA N/A · 32768 ctx

    168.62 tok/s

    Decode

    623.7 tok/s

    Prefill

    0.45 s

    TTFT

    4.42 GB

    VRAM

    —

    MTP

    16.7 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=102 running model_id=55 at Q3_K_S